search

quic / qidk

Other
76 stars 16 forks source link

VisionSolution4-PoseEstimation： How to do post-processing #8

Closed Tom-plus closed 5 months ago

Tom-plus commented 6 months ago

According to the tutorial, I have converted yolo_nas_s.onnx into DLC and quantified it. Then, I deduced the final DLC and obtained two outputs, one with shape of (1, 2100 80) and other with the shape of (1, 2100, 4) which I assume the shapes of scores and boxes respectively. My question is regarding processing these outputs. I cannot understand the Java and C++code in the tutorial. Is there a Python related tutorial available？

Tom-plus commented 6 months ago

Now that I have done the post-processing for YOLO, how should I do the post-processing for HRNET?

quic-rneti commented 6 months ago

We already have pre/post for YoloNAS, and HRNET in the application in C++. Are you looking for python pre/post ?

quic-rneti commented 6 months ago

Python pre/post for YoloNAS available here - https://github.com/quic/ai-stack-models/tree/main/models-for-solutions/03-object-detection/yolonas

For HRNet, we have it internally, but did not public to public repository, yet.

quic-rneti commented 5 months ago

Closing, as there is no further update