Tracking using provided detections

[EddieEduardo]

Hello, happy new year! Thanks for sharing the codes.

Can I use these codes to track my own detections? Because I have obtained the detections from the front-end detector, so I just want to use the codes to do the back-end tracking (associating), the size of my images is all 450x705. But I found the codes are integrated with the front-end detector, could you point out the place in which I can use my detections? Thank you very much.

Looking forward to your reply.

[iliiliiliili]

Hi @EddieEduardo, you can replace detector in VideoTracker of yolov3_deepsort.py.

[susuhu]

Hi @EddieEduardo, you can replace detector in VideoTracker of yolov3_deepsort.py.

I want to use Fasterrcnn from pytorch as the detector and I've changed the detector in VideoTracker of deepsort.py, but then when it triggers Extractor in feature_extractor.py, an error occurred:

"state_dict = torch.load(model_path, map_location=lambda storage, loc: storage)[ "net_dict"] TypeError: 'FasterRCNN' object is not subscriptable

even if I load the state_dict of fasterrcnn, there is no "net_dict", I wonder what this is? And I don't quite understand the deep_sort.deep.model.py class Net. how should it be changed according to our chosen model?