ifzhang / FairMOT

[IJCV-2021] FairMOT: On the Fairness of Detection and Re-Identification in Multi-Object Tracking
MIT License
3.99k stars 935 forks source link

After Training Multi-Class and Multi-Object model, can't use demo.py #252

Closed shan-ying closed 3 years ago

shan-ying commented 4 years ago

This is the great work! Thank you for it! :)

I have a question about the Multi-Class and Multi-Object. I trained the model use the multi-class, the step include set the class number is 5 in opt (In opt.txt "num_classes: 5") and change the dataset's class id.

Now I finished training, want to predict my video, but has this error. image

I can't find the problem, could you help me to fix this problem? Thank you :)

ifzhang commented 3 years ago

During tracking, each class should be separately computed. A multi-class FairMOT repo can be found here: https://github.com/CaptainEven/MCMOT Hope that can help you!