alibaba / u2mot

[ICCV 2023] Uncertainty-aware Unsupervised Multi-Object Tracking
MIT License
21 stars 0 forks source link

How to train on unlabeled data? #3

Open siddtmb opened 5 months ago

siddtmb commented 5 months ago

How can we train on unlabeled custom data? Your instructions mentioned that we have to convert the dataset to COCO annotations, but I don't have track ID labels for my task; I have the bounding boxes from Yolo.

Please let me know. Thanks.

kail8 commented 2 months ago

Sorry for the late reply. Just set your ID labels to -1, and our code will ignore the IDs and generate the pseudo-labels automatically.