Train on other data sets

[havinggood]

Can I use train.py to train on target detection data sets, like LaSOT, GOT10k, etc

[ahatamiz]

Hi @havinggood

The boiler plate training code is from timm library directly. As long as your data structure is as follows, it should be possible:

  ├── train
  │   ├── class1
  │   │   ├── img1.jpeg
  │   │   ├── img2.jpeg
  │   │   └── ...
  │   ├── class2
  │   │   ├── img3.jpeg
  │   │   └── ...
  │   └── ...
  └── val
      ├── class1
      │   ├── img4.jpeg
      │   ├── img5.jpeg
      │   └── ...
      ├── class2
      │   ├── img6.jpeg
      │   └── ...
      └── ...