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
3.86k stars 336 forks source link

Are sparse depth map labels available for fine-tuning to get metric depth estimation？ #55

Closed Edric-star closed 4 months ago

Edric-star commented 4 months ago

Hi @LiheYoung, again loving your work! As the title mentioned, would it be ok that I use my sparse depth map labels (instead of dense) to fine-tune? My purpose is to get the estimation of my own datasets, I would really appreciate it if you could let me know.

LiheYoung commented 4 months ago

Yes, it is ok. You can simply ignore the loss in regions without labels.

Edric-star commented 4 months ago

Yes, it is ok. You can simply ignore the loss in regions without labels.

Copy that, thanks!