infercenternet.go

package main

import (
	"log"

	tf "github.com/galeone/tensorflow/tensorflow/go"
	tg "github.com/galeone/tfgo"
	"github.com/galeone/tfgo/image"
)

/*
$ saved_model_cli show --all --dir ~/TFModels/centernet_hourglass_512x512_kpts_1/ 2>/dev/null

MetaGraphDef with tag-set: 'serve' contains the following SignatureDefs:

signature_def['__saved_model_init_op']:
  The given SavedModel SignatureDef contains the following input(s):
  The given SavedModel SignatureDef contains the following output(s):
    outputs['__saved_model_init_op'] tensor_info:
        dtype: DT_INVALID
        shape: unknown_rank
        name: NoOp
  Method name is:

signature_def['serving_default']:
  The given SavedModel SignatureDef contains the following input(s):
    inputs['input_tensor'] tensor_info:
        dtype: DT_UINT8
        shape: (1, -1, -1, 3)
        name: serving_default_input_tensor:0
  The given SavedModel SignatureDef contains the following output(s):
    outputs['detection_boxes'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 100, 4)
        name: StatefulPartitionedCall:0
    outputs['detection_classes'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 100)
        name: StatefulPartitionedCall:1
    outputs['detection_keypoint_scores'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 100, 17)
        name: StatefulPartitionedCall:2
    outputs['detection_keypoints'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 100, 17, 2)
        name: StatefulPartitionedCall:3
    outputs['detection_scores'] tensor_info:
        dtype: DT_FLOAT
        shape: (1, 100)
        name: StatefulPartitionedCall:4
    outputs['num_detections'] tensor_info:
        dtype: DT_FLOAT
        shape: (1)
        name: StatefulPartitionedCall:5
  Method name is: tensorflow/serving/predict

Defined Functions:
  Function Name: '__call__'
    Option #1
      Callable with:
        Argument #1
          input_tensor: TensorSpec(shape=(1, None, None, 3), dtype=tf.uint8, name='input_tensor')
$
*/

func inferCenterNet() {
	model := tg.LoadModel("./centernet_hourglass_512x512_kpts_1", []string{"serve"}, nil)
	root := tg.NewRoot()

	rgbimg := image.Read(root, "D1.jpg", 3)
	rgbimg = rgbimg.Clone().ResizeArea(image.Size{Height: 512, Width: 512})

	imgin := tg.Batchify(root, []tf.Output{rgbimg.Value()})
	input := tg.Exec(root, []tf.Output{imgin}, nil, &tf.SessionOptions{})
	tensor, err := tf.NewTensor(input[0].Value())
	if err != nil {
		log.Fatalln(err)
	}
	results := model.Exec(
		[]tf.Output{
			model.Op("StatefulPartitionedCall", 0), // detection_boxes
			// model.Op("StatefulPartitionedCall", 1), // detection_classes
			// model.Op("StatefulPartitionedCall", 4), // detection_scores
		},
		map[tf.Output]*tf.Tensor{model.Op("serving_default_input_tensor", 0): tensor},
	)
	log.Println(results[0].Value())
}