search

EnVision-Research / Lotus

Official Implementation of Lotus: Diffusion-based Visual Foundation Model for High-quality Dense Prediction
https://lotus3d.github.io
Apache License 2.0
317 stars 15 forks source link

how to get metric depth? #13

Open Vincent630 opened 5 days ago

Vincent630 commented 5 days ago

Hi, does this work can predict metric depth ? i found infer.sh result is not metric depth,so o wonder if this work can get metric depth and if it can ,how to get it？ thank a lot

BlingHe commented 2 days ago

Thanks for your notice!

Our model predicts affine-invariant depth maps, similar to Marigold, without the scale and shift factors to convert them to metric depth. If you need metric depth, you can align our predictions with other information, such as lidar measurements or metric depth from other methods. After alignment, you can apply the obtained scale and shift factors to convert our predictions to metric depth. You can refer to this snippest for more details on the alignment process.

Best,

johndpope commented 2 days ago

breaking update https://github.com/apple/ml-depth-pro

Vincent630 commented 2 days ago

i have test depth pro,it seem can not predict the right metric depth when i offer some custome  picture

---Original--- From: "John D. @.> Date: Fri, Oct 11, 2024 21:43 PM To: @.>; Cc: @.**@.>; Subject: Re: [EnVision-Research/Lotus] how to get metric depth? (Issue #13)

breaking update https://github.com/apple/ml-depth-pro

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>