Post Processing with UltraLightFaceDetection project

Hi:

We would like to porting a face detection project : UltraLightFaceDetection

Ultra-Light-Fast-Generic-Face-Detector-1MB/tflite/

inference_test.py and TFLiteFaceDetector.py

Ultra-Light-Fast-Generic-Face-Detector-1MB/models/onnx/

version-slim-320.onnx

===============================The test file is above=============================

We write a test.py script but have no idea to dispose post processing.

We thought the _post_processing function in TFLiteFaceDetector.py might be helpful to us.

We have problems:

  1. UltraLightFaceDetecion object can't establish due to ft.lite.interpreter failed
  2. We have no idea about that inf_results[0] and inf_results[1] corresponding to the boxes and scores respectively  in _post_processing function parameter.

Maybe we get the wrong , do you have any good suggestions?

Comments

  • @Lin Hsiehcheng

    Hi Lin Hsiehcheng,

    If you need to port this project, you must understand their preprocess and postprocess, especially what to do after their ouput node output.

    I have seen the model of this project, there are some operotars that are not supported by the Kneron hardware architecture, you have to cut them off, and the cut parts are processed on the PC.

    Here is the reference link: http://doc.kneron.com/docs/#toolchain/manual/

The discussion has been closed due to inactivity. To continue with the topic, please feel free to post a new discussion.