Open karimcossentini opened 1 year ago
Did you use Colmap to generate pose data ?
There is no difference between dtu images and custom data images. And you should check the extrinsic matrix is the w2c(world2camera ) matrix, not c2w (camera2world) matrix
When testing MVSFormer, I used camera parameters and pair.txt file generated from Colmap on DTU scan_1 images, and the results were good with a full .ply file as shown below:
However, when I generated the camera parameters and pair.txt file on a custom dataset, which consisted of 49 images of an object I took from different angles, the resulting .ply file was empty (0.00 million points are found).
the question is what's the difference between using dtu scan images and a custom data images ? Can anyone suggest why this might have happened and What steps can I take to troubleshoot the issue? thanks