Yizhi Tang

could you provide a workable model？

1

If possible， could you also provide a link to the training data？

could you provide an example of test code with a trained model?

2

Hi, thanks for your sharing of the code. I found the test_options.py in the current code, but I cannot find a test code with a trained model (only training code...

Why not use ClearModifiedPoints3D in the end of IterativeGlobalRefinement?

3

I found that ClearModifiedPoints3D() is invoked in the end of IterativeLocalRefinement, https://github.com/colmap/colmap/blob/dcfd14b300868c7eb1b360ebbca0dc21acd641b6/src/colmap/sfm/incremental_mapper.cc#L780 but it is not the same for IterativeGlobalRefinement. https://github.com/colmap/colmap/blob/dcfd14b300868c7eb1b360ebbca0dc21acd641b6/src/colmap/sfm/incremental_mapper.cc#L810 Is it expected or just a bug?

The first question is what is the purpose for users of using guess_mode. The second question is about the code related to guess_mode. In the main code, setting guess_mode=True will...

Some questions about the training code

1

### **Question 1 - the padding may be assigned in incorrect order ** During ContextCrop process, the padding is assigned as follows https://github.com/lpiccinelli-eth/UniDepth/blob/5afc0dc7a1d714ecc656b218982b0aa6ef5e8585/unidepth/datasets/pipelines/transforms.py#L1275-L1280 which seem to be in the order...

About epoch settings

1

Here, in order to solve the spawn problem, epoch is added with self.distributed_rank https://github.com/facebookresearch/vggt/blob/44b3afbd1869d8bde4894dd8ea1e293112dd5eba/training/trainer.py#L382 Here, sampler.set_epoch() will use the shifted epoch, so that different GPUs will sample at different epoches....

Depth Accurary vs. World Point Accuracy, which is better?

2

Depth can be tranform int world points with intrinsics and extrinsics, while world points can also be transform into local points or depths. But is there anyone that analyses or...