I have the following problem when running a new created dataset using colmap.
When using gsplat with this new dataset it does not reconstruct the scene, it just do random things(see attached images),
but when using gsplat with an old dataset generated from the same images it works fine.
This are the images:
OLD dataset, all seems correct
New one, seems like all is random
I have seen this message when running the training, that the scales differ to much:
NEW dataset
[Parser] 906 images, taken by 1 cameras.
Warning: COLMAP Camera is not PINHOLE. Images have distortion.
Scene scale: 157.0695753204004
Model initialized. Number of GS: 377767
OLD dataset
[Parser] 906 images, taken by 1 cameras.
Warning: COLMAP Camera is not PINHOLE. Images have distortion.
Scene scale: 2.092517541618542
Model initialized. Number of GS: 376437
I can provide any details necessary, i tried many things but not any seems to give me a clue about what is happening, the cameras.txt of this new dataset are different, but even when setting the new one to the old cameras.txt the problem continues.
I have the following problem when running a new created dataset using colmap.
When using gsplat with this new dataset it does not reconstruct the scene, it just do random things(see attached images), but when using gsplat with an old dataset generated from the same images it works fine.
This are the images:
OLD dataset, all seems correct
New one, seems like all is random
I have seen this message when running the training, that the scales differ to much:
NEW dataset
OLD dataset
I can provide any details necessary, i tried many things but not any seems to give me a clue about what is happening, the cameras.txt of this new dataset are different, but even when setting the new one to the old cameras.txt the problem continues.