mvdata 0.10.1

Gracia Dataset Convention - Python library for working with multi-view video datasets

mvdata

Python library for working with Gracia multi-view video datasets.

The package provides readers, writers, conversion tools, and GPU-aware video decode helpers for the dataset layouts documented in the docs directory.

Preview videos

Generate a compact three-view MP4 preview from any supported dataset:

from pathlib import Path
from mvdata import PreviewVideoWriter, auto_detect_dataset

dataset = auto_detect_dataset(Path("path/to/dataset"))
PreviewVideoWriter(dataset, fps=60).write("preview.mp4")

By default this uses the full dataset frame range, selects three random streams, and writes a 720px-tall 8-bit MP4. Pass frame_range=(start, end), streams=[main, top_right, bottom_right], or height=1080 to override those defaults.

Release

Prepare the next release with the local helper script:

UV_PROJECT_ENVIRONMENT=.venv312 uv run python scripts/release.py patch --dry-run
UV_PROJECT_ENVIRONMENT=.venv312 uv run python scripts/release.py patch --push

Use minor or major instead of patch for larger version bumps, or pass an explicit version with --version 1.2.3. The script updates pyproject.toml and uv.lock through uv, runs the tests and build, creates a Release vX.Y.Z commit, creates an annotated vX.Y.Z tag, and pushes the branch and tag when --push is set.

Pushing the tag triggers the GitHub Actions build, PyPI publish, and GitHub Release workflow.