How to train model on custom dataset?

[Wincioor11]

Hey can you provide any guide on how to train model on the custom dataset like crowd_human or MOT17?

I am trying to separate training for yolo and deepsort models, but I am struggling with finding a dataset format for training the deep_sort the torchvision.datasets.ImageFolder in train.py seem to require images but I would expect it to require files with detection info etc.

Thanks for your time :)

[xuarehere]

Thanks for your suggestion

The detection model and the reid model need to be trained separately

Detection model
See ./detector/ for the training method of the detection model

$  more detector/YOLOV7/README.md

Or check out yolo's official training method

Reid model training

For the training method of the reid model, see ./deep_sort/deep/train.py for more detail

$ cd ./deep_sort/deep/train.py
$ python train.py --data-dir /workspace/dataset/Market-1501/Market-1501-v15.09.15/pytorch/ --interval 10  --gpu-id 0

Fast reid
fast-reid/GETTING_STARTED.md at master · JDAI-CV/fast-reid
https://github.com/JDAI-CV/fast-reid/blob/master/GETTING_STARTED.md

Prepare your dataset

• Obeject Detection
  Train Custom Data · ultralytics/yolov5 Wiki
  https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data

• reid
  Person_reID_baseline_pytorch
  https://github.com/layumi/Person_reID_baseline_pytorch
  fast-reid/datasets at master · JDAI-CV/fast-reid
  https://github.com/JDAI-CV/fast-reid/tree/master/datasets