LiheYoung / Depth-Anything

[CVPR 2024] Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data. Foundation Model for Monocular Depth Estimation
https://depth-anything.github.io
Apache License 2.0
6.99k stars 536 forks source link

small and base weights for metric depth #227

Open shawnricecake opened 1 month ago

shawnricecake commented 1 month ago

any plan for the release of weights for small and base used for metric depth (indoor & outdoor)?

LiheYoung commented 1 month ago

Hi, you may refer to the small and base metric depth models in our Depth Anything V2: https://github.com/DepthAnything/Depth-Anything-V2/tree/main/metric_depth.

shawnricecake commented 1 month ago

Hi, you may refer to the small and base metric depth models in our Depth Anything V2: https://github.com/DepthAnything/Depth-Anything-V2/tree/main/metric_depth.

already tried, can not reproduce the accuracy

LiheYoung commented 3 weeks ago

The metric depth models provided here (in V2) are based on DPT and our pre-trained encoder. It is trained on Hypersim or Virtual KITTI 2. The setting differs from the adopted setting in our main paper for metric depth models, where we train a ZoeDepth model on NYU-D or KITTI.