How to get the error map between predicted depth and the GT depth map

nianticlabs / manydepth

[CVPR 2021] Self-supervised depth estimation from short sequences

Other

618 stars 83 forks source link

How to get the error map between predicted depth and the GT depth map #38

Open ruili3 opened 3 years ago

ruili3 commented 3 years ago

Hi,

I'm very interested in the error map visualization in Fig.4 of your paper. Do you use the projected LIDAR point cloud as the GT, or the improved ground truth image in KITTI for error computation? I wonder whether you conduct the interpolation to the GT depth map?

Can you provide the code to show the error map? Thank you a lot :D

JamieWatson683 commented 2 years ago

Hi - sorry for the delay in getting back to you!

We use the improved ground truth image for the error computation as it is more dense, and yes we interpolate using nearest neighbour to obtain more complete GT depth maps.

I'm not sure I have the code used for the paper anymore, but from memory I used skimage.morphology.dilation 3 times.

Let me know how you get on with this!

Thanks a lot.