tub-rip / event_based_optical_flow

The official implementation of "Secrets of Event-based Optical Flow" (ECCV2022 Oral and IEEE T-PAMI 2024)
GNU General Public License v3.0
138 stars 13 forks source link

Custom Dataset #17

Closed XinhuaZhang closed 1 year ago

XinhuaZhang commented 1 year ago

Hello,

I wonder if this code supports training on custom dataset? I would like to train it on the Prophesee dataset. It seems that a lot of modifications are needed.

Thanks,

shiba24 commented 1 year ago

Hi, you first need to convert data to HDF5 because currently .raw format is not supported. Can you check another library that I have been developing?: https://github.com/shiba24/event-vision-library/tree/main Example (from text to hdf5): https://github.com/shiba24/event-vision-library/blob/main/examples/convert_text_to_hdf5.py

My long-term plan is to develop the library and include some algorithms there.

shiba24 commented 1 year ago

And after that you might want to write a custom data loader class for the Psee dataset. It's relatively straightforward and simple to implement it. https://github.com/tub-rip/event_based_optical_flow/tree/main/src/data_loader Actually I have one, but am not planning to release it anytime soon.

XinhuaZhang commented 1 year ago

Hi, you first need to convert data to HDF5 because currently .raw format is not supported. Can you check another library that I have been developing?: https://github.com/shiba24/event-vision-library/tree/main Example (from text to hdf5): https://github.com/shiba24/event-vision-library/blob/main/examples/convert_text_to_hdf5.py

My long-term plan is to develop the library and include some algorithms there.

Thank you. I will try it.