search

ifzhang / FairMOT

[IJCV-2021] FairMOT: On the Fairness of Detection and Re-Identification in Multi-Object Tracking
MIT License
3.96k stars 931 forks source link

Negative MOTA score #173

Open 123mw123 opened 4 years ago

123mw123 commented 4 years ago

I trained FairMOT on custom dataset. While running track.py I am getting negative MOTA score. What does negative MOTA score mean?

IDF1 IDP IDR Rcll Prcn GT MT PT ML FP FN IDs FM MOTA MOTP IDt IDa IDm data1 0.9% 0.9% 0.9% 1.0% 1.0% 9 0 0 9 1782 1848 1 1 -94.6% 0.450 0 1 0 OVERALL 0.9% 0.9% 0.9% 1.0% 1.0% 9 0 0 9 1782 1848 1 1 -94.6% 0.450 0 1 0

MOTA is -94.6%

ifzhang commented 4 years ago

The precision and recall is almost 0 and it seems almost no object is detected. You may save the tracking results and see what happens.

Kaiqi-Chen commented 3 years ago

@ramadevsaiteja ，I also trained the FairMOT on custom dataset,then I test other dataset,finallt get the result: IDF1 IDP IDR Rcll Prcn GT MT PT ML FP FN IDs FM MOTA MOTP IDt IDa IDm KITTI-13 0.0% 0.0% nan% nan% 0.0% 0 0 0 0 135 0 0 0 -inf% nan 0 0 0

I only have the FP result, but dont have the other result

samwan517 commented 3 years ago

@ramadevsaiteja ，I also trained the FairMOT on custom dataset,then I test other dataset,finallt get the result: IDF1 IDP IDR Rcll Prcn GT MT PT ML FP FN IDs FM MOTA MOTP IDt IDa IDm KITTI-13 0.0% 0.0% nan% nan% 0.0% 0 0 0 0 135 0 0 0 -inf% nan 0 0 0

I only have the FP result, but dont have the other result

I met the same problem,have you solved this problem

gmt710 commented 2 years ago

haha, have you solved ? @samwan517 @ramadevsaiteja @Ckq-Sugar

gmt710 commented 2 years ago

The gt image size may be not same with pred.