The Ai Postprocessing Wrapper is made to be used with the lib_vision_models_pp. It allows to simplify the usage of it in the use case of app using multiple types of post processing.
Add POSTPROCESS_WRAPPER_SECTION macro to allow placement of post-processing buffers in a user-defined memory section
Automatic ordering of model outputs by size and shape to improve robustness against output index changes
Yolo-D object detection support New SSD object detection support
Yunet face detection with keypoints support Compatibility with STAI API
BlazeFace face detection with keypoints support
BlazeFace post processing bug fix
Add int8 post processing support for:
- Object detection Yolo v2
- Object detection Yolo v8
- Object detection ST Yolo X
- Single pose estimation Movenet
- Multi pose estimation Yolo v8
- Semantic segmentation Deeplab v3
Add support for BlazeFace post processing
Initial Release