Accuracy decreasing after onnx to hef conversion

General
, , , ,
4

Hi @Yusuf_Ucan

If your compiled hef is not detecting anything, I have a couple of suggestions.

  1. You can share your hef and some sample images and we can run it through our PySDK and let you know if we see the same issue. If it works well through PySDK, then it is some inference code related issue.
  2. We developed a cloud compiler tool to help users with compiling YOLO models. You can compile using our tool and test. This will eliminate compiler relates issues. Please see Early Access to DeGirum Cloud Compiler