compile tensorflow inception v3 retrain graph

idata · ‎11-22-2018

Hi,

I have trained a tensorflow inception v3 model following the below steps, can anyone suggest how to convert it to a graph for ncsdk?

1.

retrain the inception model.

ref: https://github.com/googlecodelabs/tensorflow-for-poets-2

python scripts/retrain.py --output_graph=tf_files/retrained_graph.pb --output_labels=tf_files/retrained_labels.txt --image_dir=tf_files/images/

2.

remove nodes so it can be loaded with cv2.dnn.readNetFromTensorflow

ref: http://answers.opencv.org/question/175699/readnetfromtensorflow-fails-on-retrained-nn/

python ~/tensorflow/tensorflow/python/tools/optimize_for_inference.py \
  --input graph.pb \
  --output opt_graph.pb \
  --frozen_graph True \
  --input_names DecodeJpeg/contents \
  --output_names final_result

~/tensorflow/bazel-bin/tensorflow/tools/graph_transforms/transform_graph \
  --in_graph=opt_graph.pb \
  --out_graph=final_graph.pb \
  --inputs=Mul \
  --outputs=final_result \
  --transforms="remove_nodes(op=PlaceholderWithDefault) strip_unused_nodes(type=float, shape=\"1,299,299,3\") sort_by_execution_order"