ViP-DeepLab icon indicating copy to clipboard operation
ViP-DeepLab copied to clipboard

KITTI benchmark for single image depth prediction

Open ArminMasoumian opened this issue 3 years ago • 1 comments

Firstly, thank you for providing an incredible method for depth prediction. I am trying to send my results to KITTI benchmark as well, but I have some doubts, do I need to submit disparity results or estimated depths? Also, results must be 1 channel or RGB 3 channels? I really appreciate it if you help me in this regard.

ArminMasoumian avatar May 27 '21 22:05 ArminMasoumian

KITTI depth benchmark is expecting estimated depths in the same format as the ground truth depths they provide. Hope this helps.

joe-siyuan-qiao avatar May 29 '21 01:05 joe-siyuan-qiao