Created
April 22, 2020 02:05
-
-
Save siahuat0727/617ce3a0b3ffc8be6885bc7d2f9f40b1 to your computer and use it in GitHub Desktop.
darkscnn lane detection
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: "darknet-16c-16x-3d multitask TEST 960x384, offset L3:440, L4: 312, RM DET" | |
| # SCNN part: kernel size 5, only Up-Down direction | |
| ###################### LANE #######################3 | |
| layer { | |
| name: "input" | |
| type: "Input" | |
| top: "data" | |
| input_param { | |
| shape { | |
| dim: 1 | |
| dim: 3 | |
| dim: 480 | |
| dim: 640 | |
| } | |
| } | |
| } | |
| layer{ | |
| name: "scale_data_lane" | |
| type: "Power" | |
| bottom: "data" | |
| top: "scale_data_lane" | |
| power_param { | |
| power: 1 | |
| scale: 0.00392157 | |
| shift: 0 | |
| } | |
| propagate_down: false | |
| } | |
| ########################################################## | |
| layer { | |
| name: "conv1" | |
| type: "Convolution" | |
| bottom: "scale_data_lane" | |
| top: "conv1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 16 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv1_bn" | |
| type: "BatchNorm" | |
| bottom: "conv1" | |
| top: "conv1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv1_scale" | |
| type: "Scale" | |
| bottom: "conv1" | |
| top: "conv1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv1_relu" | |
| type: "ReLU" | |
| bottom: "conv1" | |
| top: "conv1" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "pool1" | |
| type: "Pooling" | |
| bottom: "conv1" | |
| top: "pool1" | |
| pooling_param { | |
| pool: MAX | |
| kernel_size: 2 | |
| pad: 0 | |
| stride: 2 | |
| } | |
| } | |
| layer { | |
| name: "conv2" | |
| type: "Convolution" | |
| bottom: "pool1" | |
| top: "conv2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 32 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv2_bn" | |
| type: "BatchNorm" | |
| bottom: "conv2" | |
| top: "conv2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv2_scale" | |
| type: "Scale" | |
| bottom: "conv2" | |
| top: "conv2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv2_relu" | |
| type: "ReLU" | |
| bottom: "conv2" | |
| top: "conv2" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "pool2" | |
| type: "Pooling" | |
| bottom: "conv2" | |
| top: "pool2" | |
| pooling_param { | |
| pool: MAX | |
| kernel_size: 2 | |
| pad: 0 | |
| stride: 2 | |
| } | |
| } | |
| layer { | |
| name: "conv3_1" | |
| type: "Convolution" | |
| bottom: "pool2" | |
| top: "conv3_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv3_1_bn" | |
| type: "BatchNorm" | |
| bottom: "conv3_1" | |
| top: "conv3_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv3_1_scale" | |
| type: "Scale" | |
| bottom: "conv3_1" | |
| top: "conv3_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv3_1_relu" | |
| type: "ReLU" | |
| bottom: "conv3_1" | |
| top: "conv3_1" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv3_2" | |
| type: "Convolution" | |
| bottom: "conv3_1" | |
| top: "conv3_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 32 | |
| bias_term: false | |
| pad: 0 | |
| kernel_size: 1 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv3_2_bn" | |
| type: "BatchNorm" | |
| bottom: "conv3_2" | |
| top: "conv3_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv3_2_scale" | |
| type: "Scale" | |
| bottom: "conv3_2" | |
| top: "conv3_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv3_2_relu" | |
| type: "ReLU" | |
| bottom: "conv3_2" | |
| top: "conv3_2" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv3_3" | |
| type: "Convolution" | |
| bottom: "conv3_2" | |
| top: "conv3_3" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv3_3_bn" | |
| type: "BatchNorm" | |
| bottom: "conv3_3" | |
| top: "conv3_3" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv3_3_scale" | |
| type: "Scale" | |
| bottom: "conv3_3" | |
| top: "conv3_3" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv3_3_relu" | |
| type: "ReLU" | |
| bottom: "conv3_3" | |
| top: "conv3_3" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "pool3" | |
| type: "Pooling" | |
| bottom: "conv3_3" | |
| top: "pool3" | |
| pooling_param { | |
| pool: MAX | |
| kernel_size: 2 | |
| pad: 0 | |
| stride: 2 | |
| } | |
| } | |
| layer { | |
| name: "conv4_1" | |
| type: "Convolution" | |
| bottom: "pool3" | |
| top: "conv4_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 128 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv4_1_bn" | |
| type: "BatchNorm" | |
| bottom: "conv4_1" | |
| top: "conv4_1" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv4_1_scale" | |
| type: "Scale" | |
| bottom: "conv4_1" | |
| top: "conv4_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv4_1_relu" | |
| type: "ReLU" | |
| bottom: "conv4_1" | |
| top: "conv4_1" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv4_2" | |
| type: "Convolution" | |
| bottom: "conv4_1" | |
| top: "conv4_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| bias_term: false | |
| pad: 0 | |
| kernel_size: 1 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv4_2_bn" | |
| type: "BatchNorm" | |
| bottom: "conv4_2" | |
| top: "conv4_2" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv4_2_scale" | |
| type: "Scale" | |
| bottom: "conv4_2" | |
| top: "conv4_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv4_2_relu" | |
| type: "ReLU" | |
| bottom: "conv4_2" | |
| top: "conv4_2" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv4_3" | |
| type: "Convolution" | |
| bottom: "conv4_2" | |
| top: "conv4_3" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 128 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv4_3_bn" | |
| type: "BatchNorm" | |
| bottom: "conv4_3" | |
| top: "conv4_3" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv4_3_scale" | |
| type: "Scale" | |
| bottom: "conv4_3" | |
| top: "conv4_3" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv4_3_relu" | |
| type: "ReLU" | |
| bottom: "conv4_3" | |
| top: "conv4_3" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "pool4" | |
| type: "Pooling" | |
| bottom: "conv4_3" | |
| top: "pool4" | |
| pooling_param { | |
| pool: MAX | |
| kernel_size: 2 | |
| pad: 0 | |
| stride: 2 | |
| } | |
| } | |
| layer { | |
| name: "conv5_1" | |
| type: "Convolution" | |
| bottom: "pool4" | |
| top: "conv5_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 256 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv5_1_bn" | |
| type: "BatchNorm" | |
| bottom: "conv5_1" | |
| top: "conv5_1" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv5_1_scale" | |
| type: "Scale" | |
| bottom: "conv5_1" | |
| top: "conv5_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv5_1_relu" | |
| type: "ReLU" | |
| bottom: "conv5_1" | |
| top: "conv5_1" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv5_2" | |
| type: "Convolution" | |
| bottom: "conv5_1" | |
| top: "conv5_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 128 | |
| bias_term: false | |
| pad: 0 | |
| kernel_size: 1 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv5_2_bn" | |
| type: "BatchNorm" | |
| bottom: "conv5_2" | |
| top: "conv5_2" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv5_2_scale" | |
| type: "Scale" | |
| bottom: "conv5_2" | |
| top: "conv5_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv5_2_relu" | |
| type: "ReLU" | |
| bottom: "conv5_2" | |
| top: "conv5_2" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv5_3" | |
| type: "Convolution" | |
| bottom: "conv5_2" | |
| top: "conv5_3" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 256 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv5_3_bn" | |
| type: "BatchNorm" | |
| bottom: "conv5_3" | |
| top: "conv5_3" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv5_3_scale" | |
| type: "Scale" | |
| bottom: "conv5_3" | |
| top: "conv5_3" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv5_3_relu" | |
| type: "ReLU" | |
| bottom: "conv5_3" | |
| top: "conv5_3" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv5_4" | |
| type: "Convolution" | |
| bottom: "conv5_3" | |
| top: "conv5_4" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 128 | |
| bias_term: false | |
| pad: 0 | |
| kernel_size: 1 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv5_4_bn" | |
| type: "BatchNorm" | |
| bottom: "conv5_4" | |
| top: "conv5_4" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv5_4_scale" | |
| type: "Scale" | |
| bottom: "conv5_4" | |
| top: "conv5_4" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv5_4_relu" | |
| type: "ReLU" | |
| bottom: "conv5_4" | |
| top: "conv5_4" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv5_5" | |
| type: "Convolution" | |
| bottom: "conv5_4" | |
| top: "conv5_5" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 256 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv5_5_bn" | |
| type: "BatchNorm" | |
| bottom: "conv5_5" | |
| top: "conv5_5" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv5_5_scale" | |
| type: "Scale" | |
| bottom: "conv5_5" | |
| top: "conv5_5" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv5_5_relu" | |
| type: "ReLU" | |
| bottom: "conv5_5" | |
| top: "conv5_5" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "pool5" | |
| type: "Pooling" | |
| bottom: "conv5_5" | |
| top: "pool5" | |
| pooling_param { | |
| pool: MAX | |
| kernel_size: 3 | |
| pad: 1 | |
| stride: 1 | |
| } | |
| } | |
| layer { | |
| name: "conv6_1" | |
| type: "Convolution" | |
| bottom: "pool5" | |
| top: "conv6_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 512 | |
| bias_term: false | |
| pad: 2 | |
| kernel_size: 3 | |
| dilation: 2 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv6_1_bn" | |
| type: "BatchNorm" | |
| bottom: "conv6_1" | |
| top: "conv6_1" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv6_1_scale" | |
| type: "Scale" | |
| bottom: "conv6_1" | |
| top: "conv6_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv6_1_relu" | |
| type: "ReLU" | |
| bottom: "conv6_1" | |
| top: "conv6_1" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv6_2" | |
| type: "Convolution" | |
| bottom: "conv6_1" | |
| top: "conv6_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 256 | |
| bias_term: false | |
| pad: 0 | |
| kernel_size: 1 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv6_2_bn" | |
| type: "BatchNorm" | |
| bottom: "conv6_2" | |
| top: "conv6_2" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv6_2_scale" | |
| type: "Scale" | |
| bottom: "conv6_2" | |
| top: "conv6_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv6_2_relu" | |
| type: "ReLU" | |
| bottom: "conv6_2" | |
| top: "conv6_2" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv6_3" | |
| type: "Convolution" | |
| bottom: "conv6_2" | |
| top: "conv6_3" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 512 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv6_3_bn" | |
| type: "BatchNorm" | |
| bottom: "conv6_3" | |
| top: "conv6_3" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv6_3_scale" | |
| type: "Scale" | |
| bottom: "conv6_3" | |
| top: "conv6_3" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv6_3_relu" | |
| type: "ReLU" | |
| bottom: "conv6_3" | |
| top: "conv6_3" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv6_4" | |
| type: "Convolution" | |
| bottom: "conv6_3" | |
| top: "conv6_4" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 256 | |
| bias_term: false | |
| pad: 0 | |
| kernel_size: 1 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv6_4_bn" | |
| type: "BatchNorm" | |
| bottom: "conv6_4" | |
| top: "conv6_4" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv6_4_scale" | |
| type: "Scale" | |
| bottom: "conv6_4" | |
| top: "conv6_4" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv6_4_relu" | |
| type: "ReLU" | |
| bottom: "conv6_4" | |
| top: "conv6_4" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv6_5" | |
| type: "Convolution" | |
| bottom: "conv6_4" | |
| top: "conv6_5" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 512 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv6_5_bn" | |
| type: "BatchNorm" | |
| bottom: "conv6_5" | |
| top: "conv6_5" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv6_5_scale" | |
| type: "Scale" | |
| bottom: "conv6_5" | |
| top: "conv6_5" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv6_5_relu" | |
| type: "ReLU" | |
| bottom: "conv6_5" | |
| top: "conv6_5" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv7_1" | |
| type: "Convolution" | |
| bottom: "conv6_5" | |
| top: "conv7_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 512 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv7_1_bn" | |
| type: "BatchNorm" | |
| bottom: "conv7_1" | |
| top: "conv7_1" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv7_1_scale" | |
| type: "Scale" | |
| bottom: "conv7_1" | |
| top: "conv7_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv7_1_relu" | |
| type: "ReLU" | |
| bottom: "conv7_1" | |
| top: "conv7_1" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv7_2" | |
| type: "Convolution" | |
| bottom: "conv7_1" | |
| top: "conv7_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 512 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv7_2_bn" | |
| type: "BatchNorm" | |
| bottom: "conv7_2" | |
| top: "conv7_2" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv7_2_scale" | |
| type: "Scale" | |
| bottom: "conv7_2" | |
| top: "conv7_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv7_2_relu" | |
| type: "ReLU" | |
| bottom: "conv7_2" | |
| top: "conv7_2" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "concat8" | |
| type: "Concat" | |
| bottom: "conv5_5" | |
| bottom: "conv7_2" | |
| top: "concat8" | |
| concat_param { | |
| axis: 1 | |
| } | |
| } | |
| ############## slice blob according to batch size ############## | |
| ############## Parsing ############################### | |
| ######### upsample 1 ######## | |
| layer { | |
| name: "reduce1_lane" | |
| type: "Convolution" | |
| bottom: "concat8" | |
| top: "reduce1_lane" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 128 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "reduce1_lane_bn" | |
| type: "BatchNorm" | |
| bottom: "reduce1_lane" | |
| top: "reduce1_lane" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "reduce1_lane_scale" | |
| type: "Scale" | |
| bottom: "reduce1_lane" | |
| top: "reduce1_lane" | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "reduce1_lane_relu" | |
| type: "ReLU" | |
| bottom: "reduce1_lane" | |
| top: "reduce1_lane" | |
| relu_param { | |
| negative_slope: 0 | |
| } | |
| } | |
| layer { | |
| name: "deconv1_lane" | |
| type: "Deconvolution" | |
| bottom: "reduce1_lane" | |
| top: "deconv1_lane" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| kernel_size: 2 # {{2 * factor _ factor % 2}} 2 * 2 _ 0 | |
| stride: 2 # {{factor}} | |
| num_output: 64 # {{C}} | |
| #group: 48 # {{C}} | |
| pad: 0 # {{ceil((factor _ 1) / 2.)}} 2 _ 1 / 2 | |
| weight_filler: {type: "xavier" } | |
| bias_filler { | |
| type: "constant" | |
| value: 0 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "deconv1_lane_bn" | |
| type: "BatchNorm" | |
| bottom: "deconv1_lane" | |
| top: "deconv1_lane" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "deconv1_lane_scale" | |
| type: "Scale" | |
| bottom: "deconv1_lane" | |
| top: "deconv1_lane" | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "deconv1_lane_relu" | |
| type: "ReLU" | |
| bottom: "deconv1_lane" | |
| top: "deconv1_lane" | |
| relu_param { | |
| negative_slope: 0 | |
| } | |
| } | |
| ########## end upsample 1 ##### | |
| ######### upsample 2 ######## | |
| layer { | |
| name: "reorg4" | |
| type: "Convolution" | |
| bottom: "conv4_3" | |
| top: "reorg4" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| bias_term: false | |
| kernel_size: 3 | |
| pad: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "reorg4_relu" | |
| type: "ReLU" | |
| bottom: "reorg4" | |
| top: "reorg4" | |
| relu_param { | |
| negative_slope: 0 | |
| } | |
| } | |
| layer { | |
| type: "Concat" | |
| name: "concat4" | |
| bottom: "reorg4" | |
| bottom: "deconv1_lane" | |
| top: "concat4" | |
| concat_param { | |
| axis: 1 | |
| } | |
| } | |
| layer { | |
| name: "reduce2_lane" | |
| type: "Convolution" | |
| bottom: "concat4" | |
| top: "reduce2_lane" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "reduce2_lane_bn" | |
| type: "BatchNorm" | |
| bottom: "reduce2_lane" | |
| top: "reduce2_lane" | |
| batch_norm_param { | |
| use_global_stats: 1 | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "reduce2_lane_scale" | |
| type: "Scale" | |
| bottom: "reduce2_lane" | |
| top: "reduce2_lane" | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "reduce2_lane_relu" | |
| type: "ReLU" | |
| bottom: "reduce2_lane" | |
| top: "reduce2_lane" | |
| relu_param { | |
| negative_slope: 0 | |
| } | |
| } | |
| layer { | |
| name: "Slice1" | |
| bottom: "reduce2_lane" | |
| type: "Slice" | |
| top: "slice1_1" | |
| top: "slice1_2" | |
| top: "slice1_3" | |
| top: "slice1_4" | |
| top: "slice1_5" | |
| top: "slice1_6" | |
| top: "slice1_7" | |
| top: "slice1_8" | |
| top: "slice1_9" | |
| top: "slice1_10" | |
| top: "slice1_11" | |
| top: "slice1_12" | |
| top: "slice1_13" | |
| top: "slice1_14" | |
| top: "slice1_15" | |
| top: "slice1_16" | |
| top: "slice1_17" | |
| top: "slice1_18" | |
| top: "slice1_19" | |
| top: "slice1_20" | |
| top: "slice1_21" | |
| top: "slice1_22" | |
| top: "slice1_23" | |
| top: "slice1_24" | |
| top: "slice1_25" | |
| top: "slice1_26" | |
| top: "slice1_27" | |
| top: "slice1_28" | |
| top: "slice1_29" | |
| top: "slice1_30" | |
| top: "slice1_31" | |
| top: "slice1_32" | |
| top: "slice1_33" | |
| top: "slice1_34" | |
| top: "slice1_35" | |
| top: "slice1_36" | |
| top: "slice1_37" | |
| top: "slice1_38" | |
| top: "slice1_39" | |
| top: "slice1_40" | |
| top: "slice1_41" | |
| top: "slice1_42" | |
| top: "slice1_43" | |
| top: "slice1_44" | |
| top: "slice1_45" | |
| top: "slice1_46" | |
| top: "slice1_47" | |
| top: "slice1_48" | |
| top: "slice1_49" | |
| top: "slice1_50" | |
| top: "slice1_51" | |
| top: "slice1_52" | |
| top: "slice1_53" | |
| top: "slice1_54" | |
| top: "slice1_55" | |
| top: "slice1_56" | |
| top: "slice1_57" | |
| top: "slice1_58" | |
| top: "slice1_59" | |
| top: "slice1_60" | |
| slice_param { | |
| axis: 2 | |
| slice_point: 1 | |
| slice_point: 2 | |
| slice_point: 3 | |
| slice_point: 4 | |
| slice_point: 5 | |
| slice_point: 6 | |
| slice_point: 7 | |
| slice_point: 8 | |
| slice_point: 9 | |
| slice_point: 10 | |
| slice_point: 11 | |
| slice_point: 12 | |
| slice_point: 13 | |
| slice_point: 14 | |
| slice_point: 15 | |
| slice_point: 16 | |
| slice_point: 17 | |
| slice_point: 18 | |
| slice_point: 19 | |
| slice_point: 20 | |
| slice_point: 21 | |
| slice_point: 22 | |
| slice_point: 23 | |
| slice_point: 24 | |
| slice_point: 25 | |
| slice_point: 26 | |
| slice_point: 27 | |
| slice_point: 28 | |
| slice_point: 29 | |
| slice_point: 30 | |
| slice_point: 31 | |
| slice_point: 32 | |
| slice_point: 33 | |
| slice_point: 34 | |
| slice_point: 35 | |
| slice_point: 36 | |
| slice_point: 37 | |
| slice_point: 38 | |
| slice_point: 39 | |
| slice_point: 40 | |
| slice_point: 41 | |
| slice_point: 42 | |
| slice_point: 43 | |
| slice_point: 44 | |
| slice_point: 45 | |
| slice_point: 46 | |
| slice_point: 47 | |
| slice_point: 48 | |
| slice_point: 49 | |
| slice_point: 50 | |
| slice_point: 51 | |
| slice_point: 52 | |
| slice_point: 53 | |
| slice_point: 54 | |
| slice_point: 55 | |
| slice_point: 56 | |
| slice_point: 57 | |
| slice_point: 58 | |
| slice_point: 59 | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_1" | |
| bottom: "slice1_1" | |
| top: "SCNN_D_1/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_1/relu" | |
| bottom: "SCNN_D_1/message" | |
| top: "SCNN_D_1/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_1/sum" | |
| bottom: "SCNN_D_1/message" | |
| bottom: "slice1_2" | |
| top: "SCNN_D_2" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_2" | |
| bottom: "SCNN_D_2" | |
| top: "SCNN_D_2/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_2/relu" | |
| bottom: "SCNN_D_2/message" | |
| top: "SCNN_D_2/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_2/sum" | |
| bottom: "SCNN_D_2/message" | |
| bottom: "slice1_3" | |
| top: "SCNN_D_3" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_3" | |
| bottom: "SCNN_D_3" | |
| top: "SCNN_D_3/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_3/relu" | |
| bottom: "SCNN_D_3/message" | |
| top: "SCNN_D_3/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_3/sum" | |
| bottom: "SCNN_D_3/message" | |
| bottom: "slice1_4" | |
| top: "SCNN_D_4" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_4" | |
| bottom: "SCNN_D_4" | |
| top: "SCNN_D_4/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_4/relu" | |
| bottom: "SCNN_D_4/message" | |
| top: "SCNN_D_4/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_4/sum" | |
| bottom: "SCNN_D_4/message" | |
| bottom: "slice1_5" | |
| top: "SCNN_D_5" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_5" | |
| bottom: "SCNN_D_5" | |
| top: "SCNN_D_5/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_5/relu" | |
| bottom: "SCNN_D_5/message" | |
| top: "SCNN_D_5/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_5/sum" | |
| bottom: "SCNN_D_5/message" | |
| bottom: "slice1_6" | |
| top: "SCNN_D_6" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_6" | |
| bottom: "SCNN_D_6" | |
| top: "SCNN_D_6/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_6/relu" | |
| bottom: "SCNN_D_6/message" | |
| top: "SCNN_D_6/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_6/sum" | |
| bottom: "SCNN_D_6/message" | |
| bottom: "slice1_7" | |
| top: "SCNN_D_7" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_7" | |
| bottom: "SCNN_D_7" | |
| top: "SCNN_D_7/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_7/relu" | |
| bottom: "SCNN_D_7/message" | |
| top: "SCNN_D_7/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_7/sum" | |
| bottom: "SCNN_D_7/message" | |
| bottom: "slice1_8" | |
| top: "SCNN_D_8" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_8" | |
| bottom: "SCNN_D_8" | |
| top: "SCNN_D_8/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_8/relu" | |
| bottom: "SCNN_D_8/message" | |
| top: "SCNN_D_8/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_8/sum" | |
| bottom: "SCNN_D_8/message" | |
| bottom: "slice1_9" | |
| top: "SCNN_D_9" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_9" | |
| bottom: "SCNN_D_9" | |
| top: "SCNN_D_9/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_9/relu" | |
| bottom: "SCNN_D_9/message" | |
| top: "SCNN_D_9/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_9/sum" | |
| bottom: "SCNN_D_9/message" | |
| bottom: "slice1_10" | |
| top: "SCNN_D_10" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_10" | |
| bottom: "SCNN_D_10" | |
| top: "SCNN_D_10/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_10/relu" | |
| bottom: "SCNN_D_10/message" | |
| top: "SCNN_D_10/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_10/sum" | |
| bottom: "SCNN_D_10/message" | |
| bottom: "slice1_11" | |
| top: "SCNN_D_11" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_11" | |
| bottom: "SCNN_D_11" | |
| top: "SCNN_D_11/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_11/relu" | |
| bottom: "SCNN_D_11/message" | |
| top: "SCNN_D_11/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_11/sum" | |
| bottom: "SCNN_D_11/message" | |
| bottom: "slice1_12" | |
| top: "SCNN_D_12" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_12" | |
| bottom: "SCNN_D_12" | |
| top: "SCNN_D_12/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_12/relu" | |
| bottom: "SCNN_D_12/message" | |
| top: "SCNN_D_12/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_12/sum" | |
| bottom: "SCNN_D_12/message" | |
| bottom: "slice1_13" | |
| top: "SCNN_D_13" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_13" | |
| bottom: "SCNN_D_13" | |
| top: "SCNN_D_13/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_13/relu" | |
| bottom: "SCNN_D_13/message" | |
| top: "SCNN_D_13/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_13/sum" | |
| bottom: "SCNN_D_13/message" | |
| bottom: "slice1_14" | |
| top: "SCNN_D_14" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_14" | |
| bottom: "SCNN_D_14" | |
| top: "SCNN_D_14/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_14/relu" | |
| bottom: "SCNN_D_14/message" | |
| top: "SCNN_D_14/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_14/sum" | |
| bottom: "SCNN_D_14/message" | |
| bottom: "slice1_15" | |
| top: "SCNN_D_15" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_15" | |
| bottom: "SCNN_D_15" | |
| top: "SCNN_D_15/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_15/relu" | |
| bottom: "SCNN_D_15/message" | |
| top: "SCNN_D_15/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_15/sum" | |
| bottom: "SCNN_D_15/message" | |
| bottom: "slice1_16" | |
| top: "SCNN_D_16" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_16" | |
| bottom: "SCNN_D_16" | |
| top: "SCNN_D_16/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_16/relu" | |
| bottom: "SCNN_D_16/message" | |
| top: "SCNN_D_16/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_16/sum" | |
| bottom: "SCNN_D_16/message" | |
| bottom: "slice1_17" | |
| top: "SCNN_D_17" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_17" | |
| bottom: "SCNN_D_17" | |
| top: "SCNN_D_17/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_17/relu" | |
| bottom: "SCNN_D_17/message" | |
| top: "SCNN_D_17/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_17/sum" | |
| bottom: "SCNN_D_17/message" | |
| bottom: "slice1_18" | |
| top: "SCNN_D_18" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_18" | |
| bottom: "SCNN_D_18" | |
| top: "SCNN_D_18/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_18/relu" | |
| bottom: "SCNN_D_18/message" | |
| top: "SCNN_D_18/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_18/sum" | |
| bottom: "SCNN_D_18/message" | |
| bottom: "slice1_19" | |
| top: "SCNN_D_19" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_19" | |
| bottom: "SCNN_D_19" | |
| top: "SCNN_D_19/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_19/relu" | |
| bottom: "SCNN_D_19/message" | |
| top: "SCNN_D_19/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_19/sum" | |
| bottom: "SCNN_D_19/message" | |
| bottom: "slice1_20" | |
| top: "SCNN_D_20" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_20" | |
| bottom: "SCNN_D_20" | |
| top: "SCNN_D_20/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_20/relu" | |
| bottom: "SCNN_D_20/message" | |
| top: "SCNN_D_20/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_20/sum" | |
| bottom: "SCNN_D_20/message" | |
| bottom: "slice1_21" | |
| top: "SCNN_D_21" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_21" | |
| bottom: "SCNN_D_21" | |
| top: "SCNN_D_21/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_21/relu" | |
| bottom: "SCNN_D_21/message" | |
| top: "SCNN_D_21/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_21/sum" | |
| bottom: "SCNN_D_21/message" | |
| bottom: "slice1_22" | |
| top: "SCNN_D_22" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_22" | |
| bottom: "SCNN_D_22" | |
| top: "SCNN_D_22/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_22/relu" | |
| bottom: "SCNN_D_22/message" | |
| top: "SCNN_D_22/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_22/sum" | |
| bottom: "SCNN_D_22/message" | |
| bottom: "slice1_23" | |
| top: "SCNN_D_23" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_23" | |
| bottom: "SCNN_D_23" | |
| top: "SCNN_D_23/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_23/relu" | |
| bottom: "SCNN_D_23/message" | |
| top: "SCNN_D_23/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_23/sum" | |
| bottom: "SCNN_D_23/message" | |
| bottom: "slice1_24" | |
| top: "SCNN_D_24" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_24" | |
| bottom: "SCNN_D_24" | |
| top: "SCNN_D_24/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_24/relu" | |
| bottom: "SCNN_D_24/message" | |
| top: "SCNN_D_24/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_24/sum" | |
| bottom: "SCNN_D_24/message" | |
| bottom: "slice1_25" | |
| top: "SCNN_D_25" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_25" | |
| bottom: "SCNN_D_25" | |
| top: "SCNN_D_25/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_25/relu" | |
| bottom: "SCNN_D_25/message" | |
| top: "SCNN_D_25/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_25/sum" | |
| bottom: "SCNN_D_25/message" | |
| bottom: "slice1_26" | |
| top: "SCNN_D_26" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_26" | |
| bottom: "SCNN_D_26" | |
| top: "SCNN_D_26/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_26/relu" | |
| bottom: "SCNN_D_26/message" | |
| top: "SCNN_D_26/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_26/sum" | |
| bottom: "SCNN_D_26/message" | |
| bottom: "slice1_27" | |
| top: "SCNN_D_27" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_27" | |
| bottom: "SCNN_D_27" | |
| top: "SCNN_D_27/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_27/relu" | |
| bottom: "SCNN_D_27/message" | |
| top: "SCNN_D_27/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_27/sum" | |
| bottom: "SCNN_D_27/message" | |
| bottom: "slice1_28" | |
| top: "SCNN_D_28" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_28" | |
| bottom: "SCNN_D_28" | |
| top: "SCNN_D_28/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_28/relu" | |
| bottom: "SCNN_D_28/message" | |
| top: "SCNN_D_28/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_28/sum" | |
| bottom: "SCNN_D_28/message" | |
| bottom: "slice1_29" | |
| top: "SCNN_D_29" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_29" | |
| bottom: "SCNN_D_29" | |
| top: "SCNN_D_29/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_29/relu" | |
| bottom: "SCNN_D_29/message" | |
| top: "SCNN_D_29/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_29/sum" | |
| bottom: "SCNN_D_29/message" | |
| bottom: "slice1_30" | |
| top: "SCNN_D_30" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_30" | |
| bottom: "SCNN_D_30" | |
| top: "SCNN_D_30/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_30/relu" | |
| bottom: "SCNN_D_30/message" | |
| top: "SCNN_D_30/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_30/sum" | |
| bottom: "SCNN_D_30/message" | |
| bottom: "slice1_31" | |
| top: "SCNN_D_31" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_31" | |
| bottom: "SCNN_D_31" | |
| top: "SCNN_D_31/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_31/relu" | |
| bottom: "SCNN_D_31/message" | |
| top: "SCNN_D_31/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_31/sum" | |
| bottom: "SCNN_D_31/message" | |
| bottom: "slice1_32" | |
| top: "SCNN_D_32" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_32" | |
| bottom: "SCNN_D_32" | |
| top: "SCNN_D_32/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_32/relu" | |
| bottom: "SCNN_D_32/message" | |
| top: "SCNN_D_32/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_32/sum" | |
| bottom: "SCNN_D_32/message" | |
| bottom: "slice1_33" | |
| top: "SCNN_D_33" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_33" | |
| bottom: "SCNN_D_33" | |
| top: "SCNN_D_33/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_33/relu" | |
| bottom: "SCNN_D_33/message" | |
| top: "SCNN_D_33/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_33/sum" | |
| bottom: "SCNN_D_33/message" | |
| bottom: "slice1_34" | |
| top: "SCNN_D_34" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_34" | |
| bottom: "SCNN_D_34" | |
| top: "SCNN_D_34/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_34/relu" | |
| bottom: "SCNN_D_34/message" | |
| top: "SCNN_D_34/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_34/sum" | |
| bottom: "SCNN_D_34/message" | |
| bottom: "slice1_35" | |
| top: "SCNN_D_35" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_35" | |
| bottom: "SCNN_D_35" | |
| top: "SCNN_D_35/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_35/relu" | |
| bottom: "SCNN_D_35/message" | |
| top: "SCNN_D_35/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_35/sum" | |
| bottom: "SCNN_D_35/message" | |
| bottom: "slice1_36" | |
| top: "SCNN_D_36" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_36" | |
| bottom: "SCNN_D_36" | |
| top: "SCNN_D_36/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_36/relu" | |
| bottom: "SCNN_D_36/message" | |
| top: "SCNN_D_36/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_36/sum" | |
| bottom: "SCNN_D_36/message" | |
| bottom: "slice1_37" | |
| top: "SCNN_D_37" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_37" | |
| bottom: "SCNN_D_37" | |
| top: "SCNN_D_37/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_37/relu" | |
| bottom: "SCNN_D_37/message" | |
| top: "SCNN_D_37/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_37/sum" | |
| bottom: "SCNN_D_37/message" | |
| bottom: "slice1_38" | |
| top: "SCNN_D_38" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_38" | |
| bottom: "SCNN_D_38" | |
| top: "SCNN_D_38/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_38/relu" | |
| bottom: "SCNN_D_38/message" | |
| top: "SCNN_D_38/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_38/sum" | |
| bottom: "SCNN_D_38/message" | |
| bottom: "slice1_39" | |
| top: "SCNN_D_39" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_39" | |
| bottom: "SCNN_D_39" | |
| top: "SCNN_D_39/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_39/relu" | |
| bottom: "SCNN_D_39/message" | |
| top: "SCNN_D_39/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_39/sum" | |
| bottom: "SCNN_D_39/message" | |
| bottom: "slice1_40" | |
| top: "SCNN_D_40" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_40" | |
| bottom: "SCNN_D_40" | |
| top: "SCNN_D_40/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_40/relu" | |
| bottom: "SCNN_D_40/message" | |
| top: "SCNN_D_40/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_40/sum" | |
| bottom: "SCNN_D_40/message" | |
| bottom: "slice1_41" | |
| top: "SCNN_D_41" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_41" | |
| bottom: "SCNN_D_41" | |
| top: "SCNN_D_41/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_41/relu" | |
| bottom: "SCNN_D_41/message" | |
| top: "SCNN_D_41/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_41/sum" | |
| bottom: "SCNN_D_41/message" | |
| bottom: "slice1_42" | |
| top: "SCNN_D_42" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_42" | |
| bottom: "SCNN_D_42" | |
| top: "SCNN_D_42/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_42/relu" | |
| bottom: "SCNN_D_42/message" | |
| top: "SCNN_D_42/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_42/sum" | |
| bottom: "SCNN_D_42/message" | |
| bottom: "slice1_43" | |
| top: "SCNN_D_43" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_43" | |
| bottom: "SCNN_D_43" | |
| top: "SCNN_D_43/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_43/relu" | |
| bottom: "SCNN_D_43/message" | |
| top: "SCNN_D_43/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_43/sum" | |
| bottom: "SCNN_D_43/message" | |
| bottom: "slice1_44" | |
| top: "SCNN_D_44" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_44" | |
| bottom: "SCNN_D_44" | |
| top: "SCNN_D_44/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_44/relu" | |
| bottom: "SCNN_D_44/message" | |
| top: "SCNN_D_44/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_44/sum" | |
| bottom: "SCNN_D_44/message" | |
| bottom: "slice1_45" | |
| top: "SCNN_D_45" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_45" | |
| bottom: "SCNN_D_45" | |
| top: "SCNN_D_45/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_45/relu" | |
| bottom: "SCNN_D_45/message" | |
| top: "SCNN_D_45/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_45/sum" | |
| bottom: "SCNN_D_45/message" | |
| bottom: "slice1_46" | |
| top: "SCNN_D_46" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_46" | |
| bottom: "SCNN_D_46" | |
| top: "SCNN_D_46/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_46/relu" | |
| bottom: "SCNN_D_46/message" | |
| top: "SCNN_D_46/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_46/sum" | |
| bottom: "SCNN_D_46/message" | |
| bottom: "slice1_47" | |
| top: "SCNN_D_47" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_47" | |
| bottom: "SCNN_D_47" | |
| top: "SCNN_D_47/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_47/relu" | |
| bottom: "SCNN_D_47/message" | |
| top: "SCNN_D_47/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_47/sum" | |
| bottom: "SCNN_D_47/message" | |
| bottom: "slice1_48" | |
| top: "SCNN_D_48" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_48" | |
| bottom: "SCNN_D_48" | |
| top: "SCNN_D_48/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_48/relu" | |
| bottom: "SCNN_D_48/message" | |
| top: "SCNN_D_48/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_48/sum" | |
| bottom: "SCNN_D_48/message" | |
| bottom: "slice1_49" | |
| top: "SCNN_D_49" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_49" | |
| bottom: "SCNN_D_49" | |
| top: "SCNN_D_49/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_49/relu" | |
| bottom: "SCNN_D_49/message" | |
| top: "SCNN_D_49/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_49/sum" | |
| bottom: "SCNN_D_49/message" | |
| bottom: "slice1_50" | |
| top: "SCNN_D_50" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_50" | |
| bottom: "SCNN_D_50" | |
| top: "SCNN_D_50/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_50/relu" | |
| bottom: "SCNN_D_50/message" | |
| top: "SCNN_D_50/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_50/sum" | |
| bottom: "SCNN_D_50/message" | |
| bottom: "slice1_51" | |
| top: "SCNN_D_51" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_51" | |
| bottom: "SCNN_D_51" | |
| top: "SCNN_D_51/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_51/relu" | |
| bottom: "SCNN_D_51/message" | |
| top: "SCNN_D_51/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_51/sum" | |
| bottom: "SCNN_D_51/message" | |
| bottom: "slice1_52" | |
| top: "SCNN_D_52" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_52" | |
| bottom: "SCNN_D_52" | |
| top: "SCNN_D_52/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_52/relu" | |
| bottom: "SCNN_D_52/message" | |
| top: "SCNN_D_52/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_52/sum" | |
| bottom: "SCNN_D_52/message" | |
| bottom: "slice1_53" | |
| top: "SCNN_D_53" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_53" | |
| bottom: "SCNN_D_53" | |
| top: "SCNN_D_53/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_53/relu" | |
| bottom: "SCNN_D_53/message" | |
| top: "SCNN_D_53/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_53/sum" | |
| bottom: "SCNN_D_53/message" | |
| bottom: "slice1_54" | |
| top: "SCNN_D_54" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_54" | |
| bottom: "SCNN_D_54" | |
| top: "SCNN_D_54/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_54/relu" | |
| bottom: "SCNN_D_54/message" | |
| top: "SCNN_D_54/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_54/sum" | |
| bottom: "SCNN_D_54/message" | |
| bottom: "slice1_55" | |
| top: "SCNN_D_55" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_55" | |
| bottom: "SCNN_D_55" | |
| top: "SCNN_D_55/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_55/relu" | |
| bottom: "SCNN_D_55/message" | |
| top: "SCNN_D_55/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_55/sum" | |
| bottom: "SCNN_D_55/message" | |
| bottom: "slice1_56" | |
| top: "SCNN_D_56" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_56" | |
| bottom: "SCNN_D_56" | |
| top: "SCNN_D_56/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_56/relu" | |
| bottom: "SCNN_D_56/message" | |
| top: "SCNN_D_56/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_56/sum" | |
| bottom: "SCNN_D_56/message" | |
| bottom: "slice1_57" | |
| top: "SCNN_D_57" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_57" | |
| bottom: "SCNN_D_57" | |
| top: "SCNN_D_57/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_57/relu" | |
| bottom: "SCNN_D_57/message" | |
| top: "SCNN_D_57/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_57/sum" | |
| bottom: "SCNN_D_57/message" | |
| bottom: "slice1_58" | |
| top: "SCNN_D_58" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_58" | |
| bottom: "SCNN_D_58" | |
| top: "SCNN_D_58/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_58/relu" | |
| bottom: "SCNN_D_58/message" | |
| top: "SCNN_D_58/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_58/sum" | |
| bottom: "SCNN_D_58/message" | |
| bottom: "slice1_59" | |
| top: "SCNN_D_59" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_59" | |
| bottom: "SCNN_D_59" | |
| top: "SCNN_D_59/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_D_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_D_59/relu" | |
| bottom: "SCNN_D_59/message" | |
| top: "SCNN_D_59/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_D_59/sum" | |
| bottom: "SCNN_D_59/message" | |
| bottom: "slice1_60" | |
| top: "SCNN_D_60" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_60" | |
| bottom: "SCNN_D_60" | |
| top: "SCNN_U_60/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_60/relu" | |
| bottom: "SCNN_U_60/message" | |
| top: "SCNN_U_60/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_60/sum" | |
| bottom: "SCNN_U_60/message" | |
| bottom: "SCNN_D_59" | |
| top: "SCNN_U_59" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_59" | |
| bottom: "SCNN_U_59" | |
| top: "SCNN_U_59/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_59/relu" | |
| bottom: "SCNN_U_59/message" | |
| top: "SCNN_U_59/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_59/sum" | |
| bottom: "SCNN_U_59/message" | |
| bottom: "SCNN_D_58" | |
| top: "SCNN_U_58" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_58" | |
| bottom: "SCNN_U_58" | |
| top: "SCNN_U_58/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_58/relu" | |
| bottom: "SCNN_U_58/message" | |
| top: "SCNN_U_58/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_58/sum" | |
| bottom: "SCNN_U_58/message" | |
| bottom: "SCNN_D_57" | |
| top: "SCNN_U_57" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_57" | |
| bottom: "SCNN_U_57" | |
| top: "SCNN_U_57/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_57/relu" | |
| bottom: "SCNN_U_57/message" | |
| top: "SCNN_U_57/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_57/sum" | |
| bottom: "SCNN_U_57/message" | |
| bottom: "SCNN_D_56" | |
| top: "SCNN_U_56" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_56" | |
| bottom: "SCNN_U_56" | |
| top: "SCNN_U_56/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_56/relu" | |
| bottom: "SCNN_U_56/message" | |
| top: "SCNN_U_56/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_56/sum" | |
| bottom: "SCNN_U_56/message" | |
| bottom: "SCNN_D_55" | |
| top: "SCNN_U_55" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_55" | |
| bottom: "SCNN_U_55" | |
| top: "SCNN_U_55/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_55/relu" | |
| bottom: "SCNN_U_55/message" | |
| top: "SCNN_U_55/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_55/sum" | |
| bottom: "SCNN_U_55/message" | |
| bottom: "SCNN_D_54" | |
| top: "SCNN_U_54" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_54" | |
| bottom: "SCNN_U_54" | |
| top: "SCNN_U_54/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_54/relu" | |
| bottom: "SCNN_U_54/message" | |
| top: "SCNN_U_54/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_54/sum" | |
| bottom: "SCNN_U_54/message" | |
| bottom: "SCNN_D_53" | |
| top: "SCNN_U_53" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_53" | |
| bottom: "SCNN_U_53" | |
| top: "SCNN_U_53/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_53/relu" | |
| bottom: "SCNN_U_53/message" | |
| top: "SCNN_U_53/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_53/sum" | |
| bottom: "SCNN_U_53/message" | |
| bottom: "SCNN_D_52" | |
| top: "SCNN_U_52" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_52" | |
| bottom: "SCNN_U_52" | |
| top: "SCNN_U_52/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_52/relu" | |
| bottom: "SCNN_U_52/message" | |
| top: "SCNN_U_52/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_52/sum" | |
| bottom: "SCNN_U_52/message" | |
| bottom: "SCNN_D_51" | |
| top: "SCNN_U_51" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_51" | |
| bottom: "SCNN_U_51" | |
| top: "SCNN_U_51/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_51/relu" | |
| bottom: "SCNN_U_51/message" | |
| top: "SCNN_U_51/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_51/sum" | |
| bottom: "SCNN_U_51/message" | |
| bottom: "SCNN_D_50" | |
| top: "SCNN_U_50" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_50" | |
| bottom: "SCNN_U_50" | |
| top: "SCNN_U_50/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_50/relu" | |
| bottom: "SCNN_U_50/message" | |
| top: "SCNN_U_50/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_50/sum" | |
| bottom: "SCNN_U_50/message" | |
| bottom: "SCNN_D_49" | |
| top: "SCNN_U_49" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_49" | |
| bottom: "SCNN_U_49" | |
| top: "SCNN_U_49/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_49/relu" | |
| bottom: "SCNN_U_49/message" | |
| top: "SCNN_U_49/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_49/sum" | |
| bottom: "SCNN_U_49/message" | |
| bottom: "SCNN_D_48" | |
| top: "SCNN_U_48" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_48" | |
| bottom: "SCNN_U_48" | |
| top: "SCNN_U_48/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_48/relu" | |
| bottom: "SCNN_U_48/message" | |
| top: "SCNN_U_48/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_48/sum" | |
| bottom: "SCNN_U_48/message" | |
| bottom: "SCNN_D_47" | |
| top: "SCNN_U_47" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_47" | |
| bottom: "SCNN_U_47" | |
| top: "SCNN_U_47/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_47/relu" | |
| bottom: "SCNN_U_47/message" | |
| top: "SCNN_U_47/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_47/sum" | |
| bottom: "SCNN_U_47/message" | |
| bottom: "SCNN_D_46" | |
| top: "SCNN_U_46" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_46" | |
| bottom: "SCNN_U_46" | |
| top: "SCNN_U_46/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_46/relu" | |
| bottom: "SCNN_U_46/message" | |
| top: "SCNN_U_46/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_46/sum" | |
| bottom: "SCNN_U_46/message" | |
| bottom: "SCNN_D_45" | |
| top: "SCNN_U_45" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_45" | |
| bottom: "SCNN_U_45" | |
| top: "SCNN_U_45/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_45/relu" | |
| bottom: "SCNN_U_45/message" | |
| top: "SCNN_U_45/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_45/sum" | |
| bottom: "SCNN_U_45/message" | |
| bottom: "SCNN_D_44" | |
| top: "SCNN_U_44" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_44" | |
| bottom: "SCNN_U_44" | |
| top: "SCNN_U_44/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_44/relu" | |
| bottom: "SCNN_U_44/message" | |
| top: "SCNN_U_44/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_44/sum" | |
| bottom: "SCNN_U_44/message" | |
| bottom: "SCNN_D_43" | |
| top: "SCNN_U_43" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_43" | |
| bottom: "SCNN_U_43" | |
| top: "SCNN_U_43/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_43/relu" | |
| bottom: "SCNN_U_43/message" | |
| top: "SCNN_U_43/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_43/sum" | |
| bottom: "SCNN_U_43/message" | |
| bottom: "SCNN_D_42" | |
| top: "SCNN_U_42" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_42" | |
| bottom: "SCNN_U_42" | |
| top: "SCNN_U_42/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_42/relu" | |
| bottom: "SCNN_U_42/message" | |
| top: "SCNN_U_42/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_42/sum" | |
| bottom: "SCNN_U_42/message" | |
| bottom: "SCNN_D_41" | |
| top: "SCNN_U_41" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_41" | |
| bottom: "SCNN_U_41" | |
| top: "SCNN_U_41/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_41/relu" | |
| bottom: "SCNN_U_41/message" | |
| top: "SCNN_U_41/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_41/sum" | |
| bottom: "SCNN_U_41/message" | |
| bottom: "SCNN_D_40" | |
| top: "SCNN_U_40" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_40" | |
| bottom: "SCNN_U_40" | |
| top: "SCNN_U_40/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_40/relu" | |
| bottom: "SCNN_U_40/message" | |
| top: "SCNN_U_40/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_40/sum" | |
| bottom: "SCNN_U_40/message" | |
| bottom: "SCNN_D_39" | |
| top: "SCNN_U_39" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_39" | |
| bottom: "SCNN_U_39" | |
| top: "SCNN_U_39/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_39/relu" | |
| bottom: "SCNN_U_39/message" | |
| top: "SCNN_U_39/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_39/sum" | |
| bottom: "SCNN_U_39/message" | |
| bottom: "SCNN_D_38" | |
| top: "SCNN_U_38" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_38" | |
| bottom: "SCNN_U_38" | |
| top: "SCNN_U_38/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_38/relu" | |
| bottom: "SCNN_U_38/message" | |
| top: "SCNN_U_38/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_38/sum" | |
| bottom: "SCNN_U_38/message" | |
| bottom: "SCNN_D_37" | |
| top: "SCNN_U_37" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_37" | |
| bottom: "SCNN_U_37" | |
| top: "SCNN_U_37/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_37/relu" | |
| bottom: "SCNN_U_37/message" | |
| top: "SCNN_U_37/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_37/sum" | |
| bottom: "SCNN_U_37/message" | |
| bottom: "SCNN_D_36" | |
| top: "SCNN_U_36" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_36" | |
| bottom: "SCNN_U_36" | |
| top: "SCNN_U_36/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_36/relu" | |
| bottom: "SCNN_U_36/message" | |
| top: "SCNN_U_36/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_36/sum" | |
| bottom: "SCNN_U_36/message" | |
| bottom: "SCNN_D_35" | |
| top: "SCNN_U_35" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_35" | |
| bottom: "SCNN_U_35" | |
| top: "SCNN_U_35/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_35/relu" | |
| bottom: "SCNN_U_35/message" | |
| top: "SCNN_U_35/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_35/sum" | |
| bottom: "SCNN_U_35/message" | |
| bottom: "SCNN_D_34" | |
| top: "SCNN_U_34" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_34" | |
| bottom: "SCNN_U_34" | |
| top: "SCNN_U_34/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_34/relu" | |
| bottom: "SCNN_U_34/message" | |
| top: "SCNN_U_34/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_34/sum" | |
| bottom: "SCNN_U_34/message" | |
| bottom: "SCNN_D_33" | |
| top: "SCNN_U_33" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_33" | |
| bottom: "SCNN_U_33" | |
| top: "SCNN_U_33/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_33/relu" | |
| bottom: "SCNN_U_33/message" | |
| top: "SCNN_U_33/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_33/sum" | |
| bottom: "SCNN_U_33/message" | |
| bottom: "SCNN_D_32" | |
| top: "SCNN_U_32" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_32" | |
| bottom: "SCNN_U_32" | |
| top: "SCNN_U_32/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_32/relu" | |
| bottom: "SCNN_U_32/message" | |
| top: "SCNN_U_32/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_32/sum" | |
| bottom: "SCNN_U_32/message" | |
| bottom: "SCNN_D_31" | |
| top: "SCNN_U_31" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_31" | |
| bottom: "SCNN_U_31" | |
| top: "SCNN_U_31/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_31/relu" | |
| bottom: "SCNN_U_31/message" | |
| top: "SCNN_U_31/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_31/sum" | |
| bottom: "SCNN_U_31/message" | |
| bottom: "SCNN_D_30" | |
| top: "SCNN_U_30" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_30" | |
| bottom: "SCNN_U_30" | |
| top: "SCNN_U_30/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_30/relu" | |
| bottom: "SCNN_U_30/message" | |
| top: "SCNN_U_30/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_30/sum" | |
| bottom: "SCNN_U_30/message" | |
| bottom: "SCNN_D_29" | |
| top: "SCNN_U_29" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_29" | |
| bottom: "SCNN_U_29" | |
| top: "SCNN_U_29/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_29/relu" | |
| bottom: "SCNN_U_29/message" | |
| top: "SCNN_U_29/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_29/sum" | |
| bottom: "SCNN_U_29/message" | |
| bottom: "SCNN_D_28" | |
| top: "SCNN_U_28" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_28" | |
| bottom: "SCNN_U_28" | |
| top: "SCNN_U_28/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_28/relu" | |
| bottom: "SCNN_U_28/message" | |
| top: "SCNN_U_28/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_28/sum" | |
| bottom: "SCNN_U_28/message" | |
| bottom: "SCNN_D_27" | |
| top: "SCNN_U_27" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_27" | |
| bottom: "SCNN_U_27" | |
| top: "SCNN_U_27/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_27/relu" | |
| bottom: "SCNN_U_27/message" | |
| top: "SCNN_U_27/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_27/sum" | |
| bottom: "SCNN_U_27/message" | |
| bottom: "SCNN_D_26" | |
| top: "SCNN_U_26" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_26" | |
| bottom: "SCNN_U_26" | |
| top: "SCNN_U_26/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_26/relu" | |
| bottom: "SCNN_U_26/message" | |
| top: "SCNN_U_26/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_26/sum" | |
| bottom: "SCNN_U_26/message" | |
| bottom: "SCNN_D_25" | |
| top: "SCNN_U_25" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_25" | |
| bottom: "SCNN_U_25" | |
| top: "SCNN_U_25/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_25/relu" | |
| bottom: "SCNN_U_25/message" | |
| top: "SCNN_U_25/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_25/sum" | |
| bottom: "SCNN_U_25/message" | |
| bottom: "SCNN_D_24" | |
| top: "SCNN_U_24" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_24" | |
| bottom: "SCNN_U_24" | |
| top: "SCNN_U_24/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_24/relu" | |
| bottom: "SCNN_U_24/message" | |
| top: "SCNN_U_24/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_24/sum" | |
| bottom: "SCNN_U_24/message" | |
| bottom: "SCNN_D_23" | |
| top: "SCNN_U_23" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_23" | |
| bottom: "SCNN_U_23" | |
| top: "SCNN_U_23/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_23/relu" | |
| bottom: "SCNN_U_23/message" | |
| top: "SCNN_U_23/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_23/sum" | |
| bottom: "SCNN_U_23/message" | |
| bottom: "SCNN_D_22" | |
| top: "SCNN_U_22" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_22" | |
| bottom: "SCNN_U_22" | |
| top: "SCNN_U_22/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_22/relu" | |
| bottom: "SCNN_U_22/message" | |
| top: "SCNN_U_22/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_22/sum" | |
| bottom: "SCNN_U_22/message" | |
| bottom: "SCNN_D_21" | |
| top: "SCNN_U_21" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_21" | |
| bottom: "SCNN_U_21" | |
| top: "SCNN_U_21/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_21/relu" | |
| bottom: "SCNN_U_21/message" | |
| top: "SCNN_U_21/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_21/sum" | |
| bottom: "SCNN_U_21/message" | |
| bottom: "SCNN_D_20" | |
| top: "SCNN_U_20" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_20" | |
| bottom: "SCNN_U_20" | |
| top: "SCNN_U_20/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_20/relu" | |
| bottom: "SCNN_U_20/message" | |
| top: "SCNN_U_20/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_20/sum" | |
| bottom: "SCNN_U_20/message" | |
| bottom: "SCNN_D_19" | |
| top: "SCNN_U_19" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_19" | |
| bottom: "SCNN_U_19" | |
| top: "SCNN_U_19/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_19/relu" | |
| bottom: "SCNN_U_19/message" | |
| top: "SCNN_U_19/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_19/sum" | |
| bottom: "SCNN_U_19/message" | |
| bottom: "SCNN_D_18" | |
| top: "SCNN_U_18" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_18" | |
| bottom: "SCNN_U_18" | |
| top: "SCNN_U_18/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_18/relu" | |
| bottom: "SCNN_U_18/message" | |
| top: "SCNN_U_18/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_18/sum" | |
| bottom: "SCNN_U_18/message" | |
| bottom: "SCNN_D_17" | |
| top: "SCNN_U_17" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_17" | |
| bottom: "SCNN_U_17" | |
| top: "SCNN_U_17/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_17/relu" | |
| bottom: "SCNN_U_17/message" | |
| top: "SCNN_U_17/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_17/sum" | |
| bottom: "SCNN_U_17/message" | |
| bottom: "SCNN_D_16" | |
| top: "SCNN_U_16" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_16" | |
| bottom: "SCNN_U_16" | |
| top: "SCNN_U_16/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_16/relu" | |
| bottom: "SCNN_U_16/message" | |
| top: "SCNN_U_16/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_16/sum" | |
| bottom: "SCNN_U_16/message" | |
| bottom: "SCNN_D_15" | |
| top: "SCNN_U_15" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_15" | |
| bottom: "SCNN_U_15" | |
| top: "SCNN_U_15/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_15/relu" | |
| bottom: "SCNN_U_15/message" | |
| top: "SCNN_U_15/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_15/sum" | |
| bottom: "SCNN_U_15/message" | |
| bottom: "SCNN_D_14" | |
| top: "SCNN_U_14" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_14" | |
| bottom: "SCNN_U_14" | |
| top: "SCNN_U_14/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_14/relu" | |
| bottom: "SCNN_U_14/message" | |
| top: "SCNN_U_14/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_14/sum" | |
| bottom: "SCNN_U_14/message" | |
| bottom: "SCNN_D_13" | |
| top: "SCNN_U_13" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_13" | |
| bottom: "SCNN_U_13" | |
| top: "SCNN_U_13/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_13/relu" | |
| bottom: "SCNN_U_13/message" | |
| top: "SCNN_U_13/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_13/sum" | |
| bottom: "SCNN_U_13/message" | |
| bottom: "SCNN_D_12" | |
| top: "SCNN_U_12" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_12" | |
| bottom: "SCNN_U_12" | |
| top: "SCNN_U_12/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_12/relu" | |
| bottom: "SCNN_U_12/message" | |
| top: "SCNN_U_12/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_12/sum" | |
| bottom: "SCNN_U_12/message" | |
| bottom: "SCNN_D_11" | |
| top: "SCNN_U_11" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_11" | |
| bottom: "SCNN_U_11" | |
| top: "SCNN_U_11/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_11/relu" | |
| bottom: "SCNN_U_11/message" | |
| top: "SCNN_U_11/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_11/sum" | |
| bottom: "SCNN_U_11/message" | |
| bottom: "SCNN_D_10" | |
| top: "SCNN_U_10" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_10" | |
| bottom: "SCNN_U_10" | |
| top: "SCNN_U_10/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_10/relu" | |
| bottom: "SCNN_U_10/message" | |
| top: "SCNN_U_10/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_10/sum" | |
| bottom: "SCNN_U_10/message" | |
| bottom: "SCNN_D_9" | |
| top: "SCNN_U_9" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_9" | |
| bottom: "SCNN_U_9" | |
| top: "SCNN_U_9/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_9/relu" | |
| bottom: "SCNN_U_9/message" | |
| top: "SCNN_U_9/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_9/sum" | |
| bottom: "SCNN_U_9/message" | |
| bottom: "SCNN_D_8" | |
| top: "SCNN_U_8" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_8" | |
| bottom: "SCNN_U_8" | |
| top: "SCNN_U_8/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_8/relu" | |
| bottom: "SCNN_U_8/message" | |
| top: "SCNN_U_8/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_8/sum" | |
| bottom: "SCNN_U_8/message" | |
| bottom: "SCNN_D_7" | |
| top: "SCNN_U_7" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_7" | |
| bottom: "SCNN_U_7" | |
| top: "SCNN_U_7/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_7/relu" | |
| bottom: "SCNN_U_7/message" | |
| top: "SCNN_U_7/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_7/sum" | |
| bottom: "SCNN_U_7/message" | |
| bottom: "SCNN_D_6" | |
| top: "SCNN_U_6" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_6" | |
| bottom: "SCNN_U_6" | |
| top: "SCNN_U_6/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_6/relu" | |
| bottom: "SCNN_U_6/message" | |
| top: "SCNN_U_6/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_6/sum" | |
| bottom: "SCNN_U_6/message" | |
| bottom: "SCNN_D_5" | |
| top: "SCNN_U_5" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_5" | |
| bottom: "SCNN_U_5" | |
| top: "SCNN_U_5/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_5/relu" | |
| bottom: "SCNN_U_5/message" | |
| top: "SCNN_U_5/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_5/sum" | |
| bottom: "SCNN_U_5/message" | |
| bottom: "SCNN_D_4" | |
| top: "SCNN_U_4" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_4" | |
| bottom: "SCNN_U_4" | |
| top: "SCNN_U_4/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_4/relu" | |
| bottom: "SCNN_U_4/message" | |
| top: "SCNN_U_4/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_4/sum" | |
| bottom: "SCNN_U_4/message" | |
| bottom: "SCNN_D_3" | |
| top: "SCNN_U_3" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_3" | |
| bottom: "SCNN_U_3" | |
| top: "SCNN_U_3/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_3/relu" | |
| bottom: "SCNN_U_3/message" | |
| top: "SCNN_U_3/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_3/sum" | |
| bottom: "SCNN_U_3/message" | |
| bottom: "SCNN_D_2" | |
| top: "SCNN_U_2" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_2" | |
| bottom: "SCNN_U_2" | |
| top: "SCNN_U_2/message" | |
| type: "Convolution" | |
| param { | |
| name: "SCNN_U_w" | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_h: 1 | |
| kernel_w: 5 | |
| pad_h: 0 | |
| pad_w: 2 | |
| stride: 1 | |
| bias_term: false | |
| weight_filler {type: "gaussian" std: 0.03536 } | |
| } | |
| } | |
| layer { | |
| name: "SCNN_U_2/relu" | |
| bottom: "SCNN_U_2/message" | |
| top: "SCNN_U_2/message" | |
| type: "ReLU" | |
| } | |
| layer { | |
| name: "SCNN_U_2/sum" | |
| bottom: "SCNN_U_2/message" | |
| bottom: "slice1_1" | |
| top: "SCNN_U_1" | |
| type: "Eltwise" | |
| eltwise_param { | |
| operation: SUM | |
| } | |
| } | |
| layer { | |
| name: "Concat1" | |
| type: "Concat" | |
| bottom: "SCNN_U_1" | |
| bottom: "SCNN_U_2" | |
| bottom: "SCNN_U_3" | |
| bottom: "SCNN_U_4" | |
| bottom: "SCNN_U_5" | |
| bottom: "SCNN_U_6" | |
| bottom: "SCNN_U_7" | |
| bottom: "SCNN_U_8" | |
| bottom: "SCNN_U_9" | |
| bottom: "SCNN_U_10" | |
| bottom: "SCNN_U_11" | |
| bottom: "SCNN_U_12" | |
| bottom: "SCNN_U_13" | |
| bottom: "SCNN_U_14" | |
| bottom: "SCNN_U_15" | |
| bottom: "SCNN_U_16" | |
| bottom: "SCNN_U_17" | |
| bottom: "SCNN_U_18" | |
| bottom: "SCNN_U_19" | |
| bottom: "SCNN_U_20" | |
| bottom: "SCNN_U_21" | |
| bottom: "SCNN_U_22" | |
| bottom: "SCNN_U_23" | |
| bottom: "SCNN_U_24" | |
| bottom: "SCNN_U_25" | |
| bottom: "SCNN_U_26" | |
| bottom: "SCNN_U_27" | |
| bottom: "SCNN_U_28" | |
| bottom: "SCNN_U_29" | |
| bottom: "SCNN_U_30" | |
| bottom: "SCNN_U_31" | |
| bottom: "SCNN_U_32" | |
| bottom: "SCNN_U_33" | |
| bottom: "SCNN_U_34" | |
| bottom: "SCNN_U_35" | |
| bottom: "SCNN_U_36" | |
| bottom: "SCNN_U_37" | |
| bottom: "SCNN_U_38" | |
| bottom: "SCNN_U_39" | |
| bottom: "SCNN_U_40" | |
| bottom: "SCNN_U_41" | |
| bottom: "SCNN_U_42" | |
| bottom: "SCNN_U_43" | |
| bottom: "SCNN_U_44" | |
| bottom: "SCNN_U_45" | |
| bottom: "SCNN_U_46" | |
| bottom: "SCNN_U_47" | |
| bottom: "SCNN_U_48" | |
| bottom: "SCNN_U_49" | |
| bottom: "SCNN_U_50" | |
| bottom: "SCNN_U_51" | |
| bottom: "SCNN_U_52" | |
| bottom: "SCNN_U_53" | |
| bottom: "SCNN_U_54" | |
| bottom: "SCNN_U_55" | |
| bottom: "SCNN_U_56" | |
| bottom: "SCNN_U_57" | |
| bottom: "SCNN_U_58" | |
| bottom: "SCNN_U_59" | |
| bottom: "SCNN_D_60" | |
| top: "SCNN_U" | |
| concat_param { | |
| axis: 2 | |
| } | |
| } | |
| layer { | |
| name: "conv6" | |
| type: "Convolution" | |
| bottom: "SCNN_U" | |
| top: "conv6" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 64 | |
| kernel_size: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "msra" | |
| } | |
| } | |
| } | |
| layer { | |
| name: "deconv_out" | |
| type: "Deconvolution" | |
| bottom: "conv6" | |
| top: "deconv_out" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 0 | |
| } | |
| convolution_param { | |
| kernel_size: 16 # {{2 * factor _ factor % 2}} 2 * 2 _ 0 | |
| stride: 8 # {{factor}} | |
| num_output: 32 # {{C}} | |
| #group: 64 # {{C}} | |
| pad: 4 # {{ceil((factor _ 1) / 2.)}} 2 _ 1 / 2 | |
| weight_filler: {type: "xavier" } | |
| bias_filler { | |
| type: "constant" | |
| value: 0 | |
| } | |
| } | |
| } | |
| ################## semantic segmentation output layer ################### | |
| layer { | |
| name: "conv_out" | |
| type: "Convolution" | |
| bottom: "deconv_out" | |
| top: "conv_out" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 13 | |
| kernel_size: 3 | |
| pad: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "msra" | |
| } | |
| } | |
| } | |
| layer { | |
| name: "softmax" | |
| type: "Softmax" | |
| bottom: "conv_out" | |
| top: "softmax" | |
| } | |
| ################ Vanishing point subnet ############### | |
| ###### conv layers for feature translation ####### | |
| layer { | |
| name: "conv8_1" | |
| type: "Convolution" | |
| bottom: "conv7_2" | |
| top: "conv8_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 512 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 1 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv8_1_bn" | |
| type: "BatchNorm" | |
| bottom: "conv8_1" | |
| top: "conv8_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| batch_norm_param { | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv8_1_scale" | |
| type: "Scale" | |
| bottom: "conv8_1" | |
| top: "conv8_1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv8_1_relu" | |
| type: "ReLU" | |
| bottom: "conv8_1" | |
| top: "conv8_1" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "conv8_2" | |
| type: "Convolution" | |
| bottom: "conv8_1" | |
| top: "conv8_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| convolution_param { | |
| num_output: 256 | |
| bias_term: false | |
| pad: 1 | |
| kernel_size: 3 | |
| dilation: 1 | |
| stride: 2 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "conv8_2_bn" | |
| type: "BatchNorm" | |
| bottom: "conv8_2" | |
| top: "conv8_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| batch_norm_param { | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "conv8_2_scale" | |
| type: "Scale" | |
| bottom: "conv8_2" | |
| top: "conv8_2" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "conv8_2_relu" | |
| type: "ReLU" | |
| bottom: "conv8_2" | |
| top: "conv8_2" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| ########## fc layer ############# | |
| layer { | |
| name: "fc1" | |
| type: "InnerProduct" | |
| bottom: "conv8_2" | |
| top: "fc1" | |
| # learning rate and decay multipliers for the weights | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| # learning rate and decay multipliers for the biases | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 0 | |
| } | |
| inner_product_param { | |
| num_output: 64 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| bias_filler { | |
| type: "constant" | |
| value: 0 | |
| } | |
| } | |
| } | |
| layer { | |
| name: "fc1_bn" | |
| type: "BatchNorm" | |
| bottom: "fc1" | |
| top: "fc1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| batch_norm_param { | |
| eps: 1e-06 | |
| } | |
| } | |
| layer { | |
| name: "fc1_scale" | |
| type: "Scale" | |
| bottom: "fc1" | |
| top: "fc1" | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| scale_param { | |
| filler { | |
| type: "constant" | |
| value: 1 | |
| } | |
| bias_term: true | |
| } | |
| } | |
| layer { | |
| name: "fc1_relu" | |
| type: "ReLU" | |
| bottom: "fc1" | |
| top: "fc1" | |
| relu_param { | |
| negative_slope: 0.0 | |
| } | |
| } | |
| layer { | |
| name: "fc_out" | |
| type: "InnerProduct" | |
| bottom: "fc1" | |
| top: "fc_out" | |
| # learning rate and decay multipliers for the weights | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 1 | |
| } | |
| # learning rate and decay multipliers for the biases | |
| param { | |
| lr_mult: 1 | |
| decay_mult: 0 | |
| } | |
| inner_product_param { | |
| num_output: 2 | |
| weight_filler { | |
| type: "gaussian" | |
| std: 0.01 | |
| } | |
| bias_filler { | |
| type: "constant" | |
| value: 0 | |
| } | |
| } | |
| } |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment