search

DL3DV-10K / Dataset

News: the 10k dataset is ready for download.
https://dl3dv-10k.github.io/DL3DV-10K/
Other
330 stars 5 forks source link

some images are not used for sfm reconstruction #30

Open emjay73 opened 2 months ago

emjay73 commented 2 months ago

Some images are not used for SFM reconstruction.

For example, 'frame_00223.png' is not included in the frame information in the transforms.json file of the scene 'DL3DV-ALL-960P/3K/533f672b771aa2efd4da4ffa4dac6b5d89406a8455e253c67b365df53690e312'.

Also, the reconstruction results of the scene "3K/4fbfd34fd694f51029f9affb4edbe08aabb2c89060a5575a74e928c5fe8b60a3" is only composed of 16 frames.

"2K/d26afc876a34e53738d5970d071134008b059c6a1c4b605622be621063c7e387/transforms.json" only includes 10 frames.

Is this the intended result, or is some information missing?

ShengCN commented 2 months ago

DL3DV-ALL-960P/3K/533f672b771aa2efd4da4ffa4dac6b5d89406a8455e253c67b365df53690e312 seems to be good from my side. I used the following command for testing:

python download.py --odir DL3DV-10K --subset 11K --resolution 960P --file_type images+poses --clean_cache --hash 533f672b771aa2efd4da4ffa4dac6b5d89406a8455e253c67b365df53690e312

3K/4fbfd34fd694f51029f9affb4edbe08aabb2c89060a5575a74e928c5fe8b60a3 also looks good from my side.

Maybe you can try to manually download them again?

This file 2K/d26afc876a34e53738d5970d071134008b059c6a1c4b605622be621063c7e387/transforms.json does have problem. We are working on fixing it.