autonomousvision / unimatch

[TPAMI'23] Unifying Flow, Stereo and Depth Estimation
https://haofeixu.github.io/unimatch/
MIT License
980 stars 102 forks source link

Training on custom dataset #18

Closed 26828727 closed 1 year ago

26828727 commented 1 year ago

I want to train on my custom dataset, how do I pre-processing it?

haofeixu commented 1 year ago

For which task you mean?

26828727 commented 1 year ago

For optical flow, thanks!

haofeixu commented 1 year ago

This is extremely simple:

  1. Implement a dataload for your dataset, you could refer to the KITTI example: https://github.com/autonomousvision/unimatch/blob/master/dataloader/flow/datasets.py#L278-L297
  2. Define your dataset and augmentations here: https://github.com/autonomousvision/unimatch/blob/master/dataloader/flow/datasets.py#L339
  3. Train with your dataset: https://github.com/autonomousvision/unimatch/blob/master/main_flow.py

Enjoy!

26828727 commented 1 year ago

@haofeixu ok, thank you!!!