search

TRI-ML / packnet-sfm

TRI-ML Monocular Depth Estimation Repository
https://tri-ml.github.io/packnet-sfm/
MIT License
1.24k stars 243 forks source link

How to use annoted depth maps for evaluation? #126

Closed Rodger-Huang closed 3 years ago

Rodger-Huang commented 3 years ago

Hi, thanks for your great work. I want to evaluate with the improved groundtruth depth maps. I change the depth_type in the eval_kitti.yaml from velodyne to groundtruth. https://github.com/TRI-ML/packnet-sfm/blob/c03e4bf929f202ff67819340135c53778d36047f/configs/eval_kitti.yaml#L20

I want to know if this is correct? Or should I make any other changes? Do you evaluate on the full image or on the cropped image?

VitorGuizilini-TRI commented 3 years ago

Thank you for your interest in our work! Yes, that's all you need. We evaluate the full image, unless you provide a crop to be used as mask (the standard one used for KITTI is 'garg').

Rodger-Huang commented 3 years ago

Thank you for your reply. So, if I want to evaluate the full image, should I set crop: '' instead of crop: 'garg' in the eval_kitti.yaml? https://github.com/TRI-ML/packnet-sfm/blob/c03e4bf929f202ff67819340135c53778d36047f/configs/eval_kitti.yaml#L10

VitorGuizilini-TRI commented 3 years ago

Yes, then every valid ground-truth depth pixel will be evaluated.

Rodger-Huang commented 3 years ago

Thank you for your reply.