search

google / mipnerf

Apache License 2.0
894 stars 109 forks source link

Reconstructing using colmap poses #9

Closed nishant34 closed 2 years ago

nishant34 commented 2 years ago

If we estimate the poses of the lego dataset using colmap, will it work? I tried doing the same, and tried training it using train_llff.sh but it showed inconsistent results.

jonbarron commented 2 years ago

Colmap should be able to reconstruct the poses from the lego dataset, but I definitely wouldn't use the LLFF dataloader and parameters. The LLFF loader assumes that you're looking at a "front facing" scene, which is definitely not the case for the Blender data.

You probably want to use the blender code path and data loader, and write your own modification for getting the colmap cameras into the blender format.

nishant34 commented 2 years ago

Yeah, I totally agree with you, but I don't know how to align the blender and colmap poses, which when plotted appear to be sufficiently different.