WHENetのエクスポート

docker run --rm -it --gpus all \
-v `pwd`:/home/user/workdir \
-v /tmp/.X11-unix/:/tmp/.X11-unix:rw \
--device /dev/video0:/dev/video0:mwr \
--net=host \
-e XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR \
-e DISPLAY=$DISPLAY \
--privileged \
pinto0309/whenet:latest /bin/bash

git clone https://github.com/Ascend-Research/HeadPoseEstimation-WHENet.git

cd HeadPoseEstimation-WHENet

pip3 install gdown
gdown --id 1wGrwu_5etcpuu_sLIXl9Nu0dwNc8YXIH -O yolo_v3/data/head_detect.h5

python3 demo_video.py \
--video 0 \
--display simple \
--output output.mp4

        if snapshot!=None:
            self.model.load_weights(snapshot)

        for l in self.model.layers[1:]:
            l.trainable = False

        self.model.summary()
        self.model.save('model.h5')

        import tensorflow as tf
        from keras.models import load_model
        from keras import backend as K
        from tensorflow.python.framework import graph_io
        from tensorflow.python.framework import graph_util
        def freeze_graph(session, output, save_pb_dir='.', save_pb_name='frozen_model.pb', save_pb_as_text=False):
            graph = session.graph
            with graph.as_default():
                graphdef_inf = tf.graph_util.remove_training_nodes(graph.as_graph_def())
                graphdef_frozen = graph_util.convert_variables_to_constants(session, graphdef_inf, output)
                graph_io.write_graph(graphdef_frozen, save_pb_dir, save_pb_name, as_text=save_pb_as_text)
                return graphdef_frozen
        
        K.clear_session()
        K.set_learning_phase(0)
        model = load_model('model.h5')
        session = K.get_session()
        freeze_graph(session, [out.op.name for out in model.outputs], save_pb_dir='.')

rm -rf openvino
python3 $INTEL_OPENVINO_DIR/deployment_tools/model_optimizer/mo_tf.py \
--input_model frozen_model.pb \
--input_shape [1,224,224,3] \
--output_dir openvino