Use YoloV7 detector with Byte Track

[vineetsharma14]

Firstly, thank you so much for this awesome repo !

I want to use YoloV7 Object Detector in place of YoloX detector with ByteTrack, as YoloV7 tends to perform better than YoloX is what I have observed in my particular use case.

I went through the readme where it is mentioned to use custom detectors with ByteTrack

https://github.com/ifzhang/ByteTrack#combining-byte-with-other-detectors

I also tried reverse engineering the code from the tools/demo_track.py script by placing breakpoints and observing the results of variables at different places. However, I could not find an intuitive way wherein I can pass the YoloV7's detection to the tracker.

I feel this is the place where one can swap the existing detector with the custom detector

https://github.com/ifzhang/ByteTrack/blob/72ca8b45d36caf5a39e949c6aa815d9abffd1ab5/tools/demo_track.py#L261

The format of 'outputs' variable obtained from the detection is not very clear in this line

outputs, img_info = predictor.inference(frame, timer)

Can you please guide us on how one can use YoloV7 detector in place of YoloX detector with ByteTrack.

Thanks !

[hruthiksiva]

we are also facing the same problem. Let me know if you got the solution.

[nemonameless]

YOLOv7 ByteTrack Implementation by PaddleDetection_YOLOSeries see #268