Jianyuan Wang

Hi yes, you can simply finetune the model by using an additional mlp to project the the known intrinsics to a latent vector, and then inject this latent vector to...

I never tried it before. To my best guess, you need to filter self.sequence_list based on the single_sequence_subset https://github.com/facebookresearch/vggt/blob/97bbde571faddde3ace3cfa7724a20448026c4c8/training/data/datasets/co3d.py#L154C14-L154C27

Hi, I think this is because the script converts the image width to 518 pixels. You need to scale the image size and focal length back to the original resolution...

I will provide an instruction on how to use nerfstudio over the outputs.

Hi all, I just uploaded a new file: https://github.com/facebookresearch/vggt/blob/main/demo_colmap.py The saved output is the standard colmap sparse reconstruction and can be directly fed into nerfstudio for nerf/gau training. For example,...

Hi, We do not support (1) or (2) by now. For (3), can you share more details? Is (3) relocalization?

Hi can you check if the tracking results look good to you? It is a known issue that the tracking head of the released ckpt seems not that good compared...

Hi honestly this depends on the quality of the 'pseudo ground truth' you would expect. It might be better to visually check the results on your data by eyes

Hi @ricshaw , Thanks for sharing. This is a known issue and has been reported by other users. In theory it should work but something seems to be wrong. I...

Hi @xy-gao fantastic! Really appreciated your great job! Would you mind if I share this in twitter or mention it in the github repo? I hope to make it more...