Extra bboxes when test the yolov7-tiny-pose generated TensorRT engine

[ICEY22HKU]

Hi @nanmi, thanks for your code! I has trained a yolov7-tiny-pose model and do the reparameterization. I used the pt model to do the detection and the result is correct. While after I generated the tensorrt engine through your approach, I got a test result like this:

Each target has two bboxes. It is weird and I don't know how to solve it. Could you help me to find the reason? Thanks so much!