search

ifzhang / ByteTrack

[ECCV 2022] ByteTrack: Multi-Object Tracking by Associating Every Detection Box
MIT License
4.85k stars 916 forks source link

Questions on training custom dataset #45

Open huanmx opened 3 years ago

huanmx commented 3 years ago

Thanks for your great work!

There are two questions after I trained a model on my dataset.

  1. Did you try focal loss. If I want to use focal loss, how can I modify the source code.
  2. Objects on my dataset are quite small. The result is not satisfactory. How can I improve the track results on small objects.
ImSuMyatNoe commented 3 years ago

@huanmx May I know how to train the custom dataset? As my datasets are COCO format and json annotation format. I do not have frame id and track id like MOT17 (the example in this repository). How do I create the custom dataset format? Could you please tell me how to do it as I am newbie in this field:(

Double-zh commented 3 years ago

@huanmx May I know how to train the custom dataset? As my datasets are COCO format and json annotation format. I do not have frame id and track id like MOT17 (the example in this repository). How do I create the custom dataset format? Could you please tell me how to do it as I am newbie in this field:(

I have encountered this kind of problem now, have you solved it?

ImSuMyatNoe commented 3 years ago

@Double-zh Have you solved it already?

Double-zh commented 3 years ago

@Double-zh Have you solved it already?

Hello, here can solve your problem https://github.com/Double-zh/ByteTrack

abhigoku10 commented 3 years ago

@Double-zh can we use ur repo to training custom classes also like bags, cars ??

iamrajee commented 3 years ago

@Double-zh can we use ur repo to training custom classes also like bags, cars ??

+1 I also wanted to know the same.

Double-zh commented 3 years ago

@Double-zh我们可以使用你的 repo 来训练自定义课程，比如包、汽车吗？？

yes

abhigoku10 commented 3 years ago

@Double-zh thanks for the response , the author uses yolox-X as the backbone can we use yolox-L as the backbone in ur code base ?

Double-zh commented 3 years ago

@Double-zh thanks for the response , the author uses yolox-X as the backbone can we use yolox-L as the backbone in ur code base ?

yes，All backbone[yolox - (s, m, l, x)] can be used.

abhigoku10 commented 3 years ago

@Double-zh thanks for the response do u have the trained model either of yolox s,m,l trained on 4 datasets? can you please share the model on google drive or one drive ??