I have been preprocessing datasets from distinct captures (moments in time and camera sizes). I could only acquire this by using hloc, obtaining the images.bin and cameras.bin. However there is the problem that I cannot export the transforms.json in which each frame is from a distinct camera (f.e. 18 frames for 18 cameras).
I have been preprocessing datasets from distinct captures (moments in time and camera sizes). I could only acquire this by using
hloc
, obtaining theimages.bin
andcameras.bin
. However there is the problem that I cannot export thetransforms.json
in which each frame is from a distinct camera (f.e. 18 frames for 18 cameras).This error arises https://github.com/nerfstudio-project/nerfstudio/blob/main/nerfstudio/process_data/colmap_utils.py#L461
I commented this assertion and the code works, however, it is considering only the first camera and so the output transforms.json is incorrect.
Here is the
cam_id_to_camera
:and here the
frames
:In https://github.com/autonomousvision/sdfstudio/blob/master/scripts/heritage_to_nerfstudio.py#L87 there is a workaround for this, however, it requires a config file with these parameters: radius, min_track_length, voxel_size, origin.
In https://github.com/InternLandMark/LandMark?tab=readme-ov-file#prepare-dataset define this format of
transforms.json
also for multifocal sfm, see:Already opened an issue in nerfstudio https://github.com/nerfstudio-project/nerfstudio/issues/3057