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
Andromeda:tensorflow vade$ time bazel-bin/tensorflow/examples/label_image/label_image | |
I tensorflow/core/util/stat_summarizer.cc:33] StatSummarizer found 514 nodes | |
I tensorflow/core/util/stat_summarizer.cc:353] Total time (us): curr=9316873 count=11 runs, avg 9317 ms, 514 nodes defined 504 nodes observed | |
128366.4 avg KB per run. | |
============ By run order (ms) ================= | |
[start] [first] [avg] [%] [cdf%] [Op] [Name] | |
0.000 0.086 0.086 0.001% 0.001% _SOURCE | |
0.125 0.027 0.027 0.000% 0.001% Const mixed/join/concat_dim | |
0.158 0.007 0.007 0.000% 0.001% Const pool_3/_reshape/shape |
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
time bazel-bin/tensorflow/examples/label_image/label_image | |
I tensorflow/core/util/stat_summarizer.cc:33] StatSummarizer found 1004 nodes | |
W tensorflow/core/framework/op_def_util.cc:332] Op BatchNormWithGlobalNormalization is deprecated. It will cease to work in GraphDef version 9. Use tf.nn.batch_normalization(). | |
I tensorflow/core/util/stat_summarizer.cc:353] Total time (us): curr=10922857 count=11 runs, avg 1.092e+04 ms, 1004 nodes defined 901 nodes observed | |
128366.4 avg KB per run. | |
============ By run order (ms) ================= | |
[start] [first] [avg] [%] [cdf%] [Op] [Name] | |
0.000 0.204 0.204 0.002% 0.002% _SOURCE | |
0.273 0.031 0.031 0.000% 0.002% Const mixed_9/tower/conv/batchnorm/gamma |
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
/* Copyright 2015 The TensorFlow Authors. All Rights Reserved. | |
Licensed under the Apache License, Version 2.0 (the "License"); | |
you may not use this file except in compliance with the License. | |
You may obtain a copy of the License at | |
http://www.apache.org/licenses/LICENSE-2.0 | |
Unless required by applicable law or agreed to in writing, software | |
distributed under the License is distributed on an "AS IS" BASIS, |
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
Tensorflow compiled with : bazel build -c opt --copt=-mavx --cxxopt=-fno-exceptions --cxxopt=--std=c++11 --cxxopt=-DNDEBUG --cxxopt=-DNOTFDBG --cxxopt=-O2 --cxxopt=-DUSE_GEMM_FOR_CONV //tensorflow:libtensorflow_cc.so | |
Graph : Inception V3 post running Inference Optimizer | |
Output of custom app running TF, 222 frames took 28.129690 seconds | |
I tensorflow/core/util/stat_summarizer.cc:353] Total time (us): curr=41604678 count=11 runs, avg 4.16e+04 ms, 514 nodes defined 514 nodes observed | |
28625707.2 avg KB per run. |
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
Tensorflow compiled with : bazel build -c opt --copt=-mavx --cxxopt=-fno-exceptions --cxxopt=--std=c++11 --cxxopt=-DNDEBUG --cxxopt=-DNOTFDBG --cxxopt=-O2 --cxxopt=-DUSE_GEMM_FOR_CONV //tensorflow:libtensorflow_cc.so | |
Graph : Inception V3 no alterations. | |
Output of custom app running TF, 222 frames took 32.598143 seconds | |
I tensorflow/core/util/stat_summarizer.cc:353] Total time (us): curr=48605744 count=11 runs, avg 4.861e+04 ms, 1004 nodes defined 994 nodes observed | |
28625707.2 avg KB per run. | |
============ By run order (ms) ================= |
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
Tensorflow compiled with : bazel build -c opt --copt=-mavx --cxxopt=-fno-exceptions --cxxopt=--std=c++11 --cxxopt=-DNDEBUG --cxxopt=-DNOTFDBG --cxxopt=-O2 --cxxopt=-DUSE_GEMM_FOR_CONV //tensorflow:libtensorflow_cc.so | |
Graph : Inception V3 post running inference Optimizer + Quantizer with mode eightbit | |
Output of custom app running TF, 222 frames took 63.174700 seconds | |
I tensorflow/core/util/stat_summarizer.cc:33] StatSummarizer found 1282 nodes | |
I tensorflow/core/util/stat_summarizer.cc:353] Total time (us): curr=108301617 count=11 runs, avg 1.083e+05 ms, 1282 nodes defined 1282 nodes observed | |
23220669.5 avg KB per run. |
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
Tensorflow compiled with : bazel build -c opt --copt=-mavx --cxxopt=-fno-exceptions --cxxopt=--std=c++11 --cxxopt=-DNDEBUG --cxxopt=-DNOTFDBG --cxxopt=-O2 --cxxopt=-DUSE_GEMM_FOR_CONV //tensorflow:libtensorflow_cc.so | |
Graph : Inception V3 post running inference Optimizer + Quantizer with mode weights_rounded | |
Output of custom app running TF, 222 frames took 25.201791 seconds | |
I tensorflow/core/util/stat_summarizer.cc:33] StatSummarizer found 514 nodes | |
I tensorflow/core/util/stat_summarizer.cc:353] Total time (us): curr=36817771 count=11 runs, avg 3.682e+04 ms, 514 nodes defined 514 nodes observed | |
28625707.2 avg KB per run. |
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
Retraining (ie https://www.tensorflow.org/versions/r0.11/how_tos/image_retraining/index.html ) doesnt really go into | |
nuances about what types of labels you should choose based on your model. | |
Since InceptionV3 is an object recognition task, and the penultimate layer (pool 3) contains some 2048 vector length descriptions that | |
somehow infer various 'objectness' traits, its far better to say: | |
train for labels that tend toward objectness (lamp, lampshade, chandelier, standing lamp, desk lamp) | |
than train for labels that then to abstract image features like composition: chaotic, patterned, symmetric, asymmetric, mirrored, circular, diagonal, natural (photographic) , synthetic) | |
If I were interested in the latter labeling (ie, meta-features), is it more sensible to: |
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
unique_ptr<Frame> Decoder::convertVideoFrame( const Frame &frame ) const | |
{ | |
CI_ASSERT( frame.getMediaType() == AVMEDIA_TYPE_VIDEO ); | |
unique_ptr<Frame> result( new FrameVideo( frame.getTimeBase() ) ); | |
result->getAvFrame()->format = AV_PIX_FMT_RGB24; | |
result->getAvFrame()->width = frame.getAvFrame()->width; | |
result->getAvFrame()->height = frame.getAvFrame()->height; | |
// allocate backing |
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
void Encoder::configureVideoStreams() | |
{ | |
// Preset options: | |
// ultrafast superfast veryfast faster fast medium slow slower veryslow placebo | |
// see : http://dev.beandog.org/x264_preset_reference.html | |
AVDictionary *optionsDict = NULL; | |
// careful with ultrafast - it seems to force constrained baseline?; this call also allocates | |
av_dict_set( &optionsDict, "preset","superfast", 0 ); | |