Integrate a new Pose Estimation model along with the existing tracking

[hoangtnm]

Hi, I'm trying to replace HRNet with yolov7-pose, which returns bounding boxes and poses at the same time. However, I had a problem aligning the poses to the track_bboxs.
For example, if two persons are in the video, and I run python gen_skes.py -v baseball.mp4 -np 1 --animation, I will get one pose for first person ID util that ID disappears, which is okay. In my case, if I run with -np 1 and my yolov7-pose returns two bounding boxes and two poses, how to have the same behaviour with your original code? Can you give some suggestions?