Jianyuan Wang

Hi all, I just uploaded a new file demo_colmap.py, which supports an initial version of vggt+ba (not widely tested though)

Hi it seems the points are filtered out by confidence. Please try to adjust the value here: https://github.com/facebookresearch/vggt/blob/8e3d2809c1c93fa27298c290605841d7e876b285/demo_colmap.py#L60

Hi @xiemeilong , If you are happy with the result of ```demo_gradio```, you can just run ```demo_colmap``` without --use_ba. The flag ```conf_thres_value``` will only control ```demo_colmap``` without BA.

Hi, yes, in the early stage of training, the result of the first frame will be much better. In the later stage, the result of the first frame should be...

It is normal. We also found L2 is faster for convergence but not good at training stability.

Hi, Honestly I think this might be too heavy by now. Can you have a look at https://github.com/facebookresearch/co-tracker?

Hi this is usually because the default hyper-parameters are too strict for your case. For example: https://github.com/facebookresearch/vggt/blob/cd02675241e8b6c8c7f09f211fbc33c88580da7a/demo_colmap.py#L43-L60 You can use a higher max_reproj_error, lower vis_thresh, higher query_frame_num, higher max_query_pts or...

Hi Arthur, Thanks for the suggestion! We haven’t planned for high-resolution mesh reconstruction yet, but I can discuss it with my coauthors. Could you share your thoughts on recent meshing...

Hey please check (1) if you are using the correct version of gradio (2) if python 3.12 is supported by gradio.

Hi @ricshaw , This speed looks unreasonable to me. Can you check the discussion in this issue? It is quite possible that the ~30 seconds you mentioned above included the...