search

DepthAnything / Depth-Anything-V2

[NeurIPS 2024] Depth Anything V2. A More Capable Foundation Model for Monocular Depth Estimation
https://depth-anything-v2.github.io
Apache License 2.0
3.88k stars 336 forks source link

Smoothing value used for metric to disparity scale conversion of GT depth maps #195

Open Aradhye2002 opened 2 weeks ago

Aradhye2002 commented 2 weeks ago

Hi,

I wanted to find out what is c in the below equation:

disparity_gt_depth = 1/(c + metric_gt_depth)

As per your V1 paper, you mention converting the ground truth metric depth to disarity depth before normalizing to 0~1 and then applying your loss function. I wanted to find out the exact coefficient used (c here).

Thanks

LiheYoung commented 1 week ago

Hi, we do not use c here. We simply convert the depth to disparity by disparity = 1 / depth.