TRI-ML / packnet-sfm

TRI-ML Monocular Depth Estimation Repository
https://tri-ml.github.io/packnet-sfm/
MIT License
1.21k stars 241 forks source link

The question for kitti dataset #211

Open Youyi7 opened 2 years ago

Youyi7 commented 2 years ago

Hi, for the training list you provide in eigen_train, the corresponding depth data in the first line seems not exist

2011_09_28/2011_09_28_drive_0001_sync/image_02/data/0000000105.png 2011_09_28/2011_09_28_drive_0001_sync/image_03/data/0000000105.png

I could only find image data in KITTI_raw, while there is no corresponding depth data in KITTI_depth

Do I make some mistake in your code?

Thank you very much!

Youyi7 commented 2 years ago

On the other hand, for KITTI depth, do I need to do some preprocessing to interpolate the depth map to fill in holes?

Thank you very much

VitorGuizilini-TRI commented 2 years ago

The data loader finds the corresponding depth map automatically, you don't need to provide the path. And no, you don't need to fill in holes, we support training and supervision using sparse depth maps.