search

visinf / irr

Iterative Residual Refinement for Joint Optical Flow and Occlusion Estimation (CVPR 2019)
Apache License 2.0
194 stars 32 forks source link

how to prepare data for kitti testing #35

Closed RokiaAbdeen closed 3 years ago

RokiaAbdeen commented 3 years ago

could you explain please what are these files that required to upload to kitti_2015 (Provide a zip file which contains the 'disp_0' directory (stereo), the 'flow' directory (flow), or the 'disp_0', 'disp_1' and 'flow' directories (scene flow) in its root folder. Use the file format and naming described in the readme.txt (000000_10.png,...,000199_10.png). how to save them? thanks in advance..

hurjunhwa commented 3 years ago

Hi,

You can find the information of what those files are in the text file of the development kit in KITTI benchmark. If you want to evaluate optical flow only, you can just upload the flow folder.

They are the same format as the KITTI GT data, 16 bit png. You can get it if you include --save_result_png=True in the script file. Or you can use this function, https://github.com/visinf/irr/blob/f0eba07773806941853d7cb7f7ad28307a288a20/utils/flow.py#L37-L62

Much simpler code can be found here as well: https://github.com/princeton-vl/RAFT/blob/master/core/utils/frame_utils.py

RokiaAbdeen commented 3 years ago

Thanks so much for your reply I have another problem with kitti_12, I used the same image format as kitti_15 but I got wrong image format could you explain please how to write the flow images for kitti_12 benchmark thanks again

hurjunhwa commented 3 years ago

I don't remember clearly since it's been a bit long time ago. I think it would be the same as KITTI 2015. Please read the readme file from KITTI 2012 carefully! the answer is there!

RokiaAbdeen commented 3 years ago

I have read it , I just needed the pytorch code to write the flow files the image format is not same as kitti_15, I used the same format but I got image format error but thank you anyway