Datasets:
image
imagewidth (px) 378
4k
| label
class label 5
classes |
---|---|
0edits
|
|
0edits
|
|
0edits
|
|
0edits
|
|
0edits
|
|
0edits
|
|
0edits
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
0edits
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
|
1images
|
End of preview. Expand
in Dataset Viewer.
BilaRF Dataset
Project Page | Arxiv | Code
This dataset contains our own captured nighttime scenes, synthetic data generated from RawNeRF dataset, and editing samples.
To use the data, please go to 'Files and versions' and download 'bilarf_data.zip'.
The dataset follows the file structure of NeRF LLFF data (forward-facing scenes).
In addition, editing samples are stored in the 'edits/' directory. The filename of each editing is formatted in 'editX
_color_split-name
_index
.png',
where split-name
could be 'train', 'test', 'all', or 'path', and index
is the camera index in the split.
We also provide 'ext_metadata.json' that can offer info about scenes. The data loader in our code currently supports the following two fields:
{
/** The `spiral_radius_scale` field specifies the radius of spiral camera path to
ensure view synthesis is not out of the bound of the reconstructed scene. */
"spiral_radius_scale": 0.5,
/** For scenes synthesized from RawNeRF dataset, `no_factor_suffix` is set to `true`
to suggest loading downsized training images directly from the 'images/' directory
instead of 'images_X/', where X is specified by `Config.factor`. */
"no_factor_suffix": true
}
- Downloads last month
- 2