3D model in the output?

[AbrahamLincoln33]

hi, can we have 3d reconstruction of input video in the output? like ".glb" files? (like point clouds in your github page)

[gaoxiangjun]

We don't conduct 3D recon, just depth estimation. The point clouds are visualized using "viser".

[basit-MT]

@gaoxiangjun is there a code for point cloud generation?

[gaoxiangjun]

@gaoxiangjun is there a code for point cloud generation?

I will upload a demo code for point cloud visualization next few days. Please stay tuned. :-)

[gaoxiangjun]

@basit-MT Please check visualization_pcd.py. :-)

[basit-MT]

Thanks @gaoxiangjun, works perfectly.

I had another question out of curiosity, do you think fine-tuning DepthCrafter to estimate metric Depth could give useful results? Have tried anything with the metric depth?

[gaoxiangjun]

Yeah. It might be possible, but we haven't tried it yet.