Convert our current model to a model for the Edge TPU

convert_for_edge_tpu_use.sh

python object_detection/export_tflite_ssd_graph.py \
  --pipeline_config_path="pipeline.config" \
  --trained_checkpoint_prefix="model.ckpt-500" \
  --output_directory="output" \
  --add_postprocessing_op=true

bazel run -c opt tensorflow/contrib/lite/toco:toco -- \
  --input_file=output/tflite_graph.pb \
  --output_file=output/detect.tflite \
  --input_shapes=1,300,300,3 \
  --input_arrays=normalized_input_image_tensor \
  --output_arrays='TFLite_Detection_PostProcess','TFLite_Detection_PostProcess:1','TFLite_Detection_PostProcess:2','TFLite_Detection_PostProcess:3'  \
  --inference_type=QUANTIZED_UINT8 \
  --mean_values=128 \
  --std_values=128 \
  --change_concat_input_ranges=false \
  --allow_custom_ops