The Fractal20220817 dataset is provided in RLDS format and contains monocular depth annotations generated using Video-Depth-Anything(Model: Video-Depth-Anything-V2-Small).
feature.json
{
"pythonClassName": "tensorflow_datasets.core.features.features_dict.FeaturesDict",
"featuresDict": {
"features": {
"attributes": {… See the full description on the dataset page:
https://huggingface.co/datasets/ucasmichael/fractal20220817_data.