How to cacluate the EPE for the optical flow 2012/2015 test sets? about df-net

[Yuliang-Zou]

Hi @YasserTahaSW , we don't have the ground truth label for test set, so we can only save it as required format and then upload to their server for evaluation.

[yas-ml]

oh, my bad!
But why does the test set of kitti rawdata for depth have results then?
I mean the results for the test set available here: data/kitti/test_files_eigen.txt and evaluated in eval_depth.py

[Yuliang-Zou]

Originally, KITTI does not have depth estimation task on their server. But KITTI raw has stereo information, so you can actually compute sparse depth ground truth.

[yas-ml]

now i got it! Thanks a lot @Yuliang-Zou ! :)