Hugging Face's logo

Datasets:
ZhengGuangze
/
BMVS_vlbm

License:
Dataset card Files
xet
Community
Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
jpg: binary
__key__: string
__url__: string
npz: null
to
{'npz': {'extrinsics': List(List(List(Value('float16')))), 'intrinsics': List(List(List(Value('float16')))), 'depth': List(List(Value('float16')))}, '__key__': Value('string'), '__url__': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 99, in get_rows_or_raise
                  return get_rows(
                         ^^^^^^^^^
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                         ^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/utils.py", line 77, in get_rows
                  rows_plus_one = list(itertools.islice(ds, rows_max_number + 1))
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2543, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2060, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2092, in _iter_arrow
                  pa_table = cast_table_to_features(pa_table, self.features)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2192, in cast_table_to_features
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              jpg: binary
              __key__: string
              __url__: string
              npz: null
              to
              {'npz': {'extrinsics': List(List(List(Value('float16')))), 'intrinsics': List(List(List(Value('float16')))), 'depth': List(List(Value('float16')))}, '__key__': Value('string'), '__url__': Value('string')}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

BlendedMVS to VLBM Dataset Conversion Report

This report summarizes the process, results, and statistics of converting the BlendedMVS dataset into the VLBM format (from over 200GB to ~51GB). Please note that the precision of the depth maps has been reduced to 'float16'.

1. Process Overview

The conversion was performed using the script tmp/data/bmvs/bmvs2vlbm.py, which handles the following tasks:

  • Image Extraction: Extracts original or masked images from blended_images and renames them to rgb_XXXXX.jpg to match the VLBM naming convention.
  • Camera Parameter Parsing: Parses cams/XXXXXXXX_cam.txt.
    • Intrinsics: Extracts the $3 \times 3$ matrix.
    • Extrinsics: Extracts the $4 \times 4$ world-to-camera transformation matrix.
    • Metadata: Estimates focal length and sensor size based on intrinsics to generate scene_info.json.
  • Depth Map Conversion:
    • Reads high-precision depth maps from .pfm files.
    • Converts them to float16 precision for storage efficiency.
    • Saves them in compressed .npz format as depth_XXXXX.npz.

2. Target Data Format

Each scene directory follows this structure:

scene_id/
├── annotations.npz        # Contains 'intrinsics' (N, 3, 3) and 'extrinsics' (N, 4, 4)
├── scene_info.json       # Scene metadata (focal_length, sensor_size, depth_range, etc.)
├── rgbs/
│   └── rgb_XXXXX.jpg      # Color images
└── depths/
    └── depth_XXXXX.npz    # Depth maps {'depth': (H, W) float16}

3. Dataset Statistics

Statistical results for the converted dataset (based on 502 scenes):

Metrics Statistics
Total Scenes 502
Total Frames 115,142
Avg. Frames per Scene ~229 frames
Total Storage Size 51.19 GB
RGB Images Size 25.86 GB
Depth Maps Size 25.32 GB
Annotations Size 2.79 MB

Key Features

  • Scalability: Over 115k high-quality multi-view frames.
  • Efficiency: Storage and I/O costs are significantly reduced by using float16 and .npz compression for depth maps while maintaining sufficient precision.
  • Compatibility: The data format is fully compatible with the VLBM training pipeline.

4. Paths

  • Source Data: data/bmvs/blendedmvs
  • VLBM Data: data/bmvs_vlbm
Downloads last month
5