kimberlime/tensorflowserving

README

1. Shell script

   • run the script with the following command ./infer_with_saved_model.sh
   • infer_with_saved_model.sh installs requirements, pulls a docker image, run the docker image, and does an inference.

2. Implmentation

   • Implemented export function and inference function with tensorflow serving.
   • export_frozen_to_saved_model.py exports frozen graph to SavedModel format.
   • image_example.py has function infer_with_serving_client()
     • infer_with_serving_client(image_data, url, return_elements) makes an inference with tensorflow serving by http request. The numpy array of image data shape should be (1, width, height, 3)

For the detailed implementation process, check https://github.com/Kimberlime/TensorFlowServing