search

ifzhang / FairMOT

[IJCV-2021] FairMOT: On the Fairness of Detection and Re-Identification in Multi-Object Tracking
MIT License
4.01k stars 934 forks source link

FaceMOT #235

Open zengwb-lx opened 4 years ago

zengwb-lx commented 4 years ago

I trained a face tracking model with HRnet18, and the tracking performance was good (it could be tracked back within 1S after being lost）But not fast enough in the 2070S（about 30fps）。could you give me some Suggestions for switching to a lighter network

ifzhang commented 4 years ago

You can use ResNet-34, which is much faster. The sh file of training ResNet-34 can be found in the experiments folder.

Ashwin-Ramesh2607 commented 4 years ago

@zengwbz Hey can you please suggest you're training process and scripts if possible. Also, which face tracking dataset did you use?

zengwb-lx commented 4 years ago

Face detection data and face recognition data are combined together Wider-Face and Casia-webface, using mobilenet_fpn backbone and Multi-scale inputs provide better performance

kevinwei30 commented 4 years ago

@zengwbz any chance you can release the trained face tracking model?

MyraBaba commented 3 years ago

@Ashwin-Ramesh2607 @zengwb-lx Hi is there any face tracking model ?