search

amazon-science / siam-mot

SiamMOT: Siamese Multi-Object Tracking
Apache License 2.0
478 stars 61 forks source link

only use tracker #7

Closed YunhaoDu closed 3 years ago

YunhaoDu commented 3 years ago

Hi, excellent work! But I have some questions. If I already have bboxes in the video and I only want to use the SiamMOT for tracking-by-detection, what should I do. Particularly, how to just inference with my bboxes. And, how to train SiamMOT with my bboxes and then inference. Thank you. (maybe issue #5 is the same problem with me)

bingshuai2019 commented 3 years ago

Yeah, the tracker allows to use public detection by setting INFERENCE.USE_GIVEN_DETECTIONS to be True, but there is some caveat to that.

You can follow the example over here: https://github.com/amazon-research/siam-mot/blob/main/readme/model_zoo.md#motchallenge-2017-test-public-detection

YunhaoDu commented 3 years ago

thanks for your reply. I've run the tracker with INFERENCE.USE_GIVEN_DETECTIONS=True. But I found that the output tracks'bboxes are not exactly same with the public detections. So the SiamMOT has fixed the input bboxes? How does it do that ? What should I do if I just want to use the public detections and don't want the tracker to fix them? thank you~ ^.^

bingshuai2019 commented 3 years ago

The code that adjusts the public detection can be found here: https://github.com/amazon-research/siam-mot/blob/main/siammot/modelling/roi_heads.py#L28-L32

In order to use the public detection as is, you can change the codes accordingly (e.g. use the codes within else statement in the above code) .