vid2dataset2 v0.2.0-alpha.2

vid2dataset2 Alpha 2

Public alpha release

v0.2.0-alpha.2 is a public alpha of the vid2dataset2 Rust + Tauri rewrite.
It provides a native desktop workflow and a shared command-line/core
implementation for turning video files into image datasets.

This is a prerelease. Expect breaking changes, incomplete features, and rough
edges as the project moves toward beta.

Highlights

• Native Tauri v2 desktop application with a React interface and a guided
  extraction workflow.
• Shared Rust core used by both the desktop application and the
  vid2dataset CLI.
• FFmpeg-backed frame extraction with configurable frame intervals and frame
  limits.
• Crop, resize, and color-space transforms, including explicit source and
  output crop-coordinate semantics.
• PNG, JPEG, and WebP image output with configurable filename prefixes and
  starting sequence numbers.
• JSONL and CSV manifests with source metadata, transform settings, output
  paths, file sizes, and SHA-256 hashes.
• Video inspection through ffprobe, FFmpeg health checks through doctor,
  recursive multi-video input, parallel workers, and overwrite controls.
• Interactive desktop preview with video scrubbing, time/frame seeking, and
  crop creation, movement, resizing, or explicit no-crop confirmation.
• Import of legacy vid2dataset YAML configuration files, while the original
  Python implementation remains available under legacy/ for reference.

Desktop distribution

Public release artifacts are built for:

• macOS ARM64 and x64: DMG installers and zipped .app bundles.
• Windows x64: MSI installer and Windows bundle artifacts.
• Linux x64: AppImage and deb packages.

Packaged applications include redistributable LGPL-compatible FFmpeg and
ffprobe binaries. The alpha build uses FFmpeg 8.1.2 with statically linked
zlib 1.3.2; the application does not download FFmpeg on first launch or
install it globally. Every release artifact is accompanied by SHA-256
checksums.

Download the artifact for your platform from the GitHub Releases page and
verify it against the matching SHA256SUMS file when possible. Alpha builds
may be unsigned, so macOS and Windows can display an operating-system security
warning on first launch.

CLI

The command-line interface is available from the workspace during this alpha:

cargo run -p vid2dataset -- doctor
cargo run -p vid2dataset -- inspect path/to/video.mp4
cargo run -p vid2dataset -- extract \
  --input path/to/video.mp4 \
  --output output/dataset \
  --every-n-frames 8 \
  --resize 640x640 \
  --format png \
  --manifest jsonl,csv

Use cargo run -p vid2dataset -- profiles import to migrate a legacy YAML
configuration into the new job model. Run presets list to inspect the
available built-in presets.

Known limitations

• This is an alpha release and is not yet a signed stable production release.
• Hosted web/server processing is not included.
• COCO and YOLO annotation exports are deferred; this release writes raw image
  outputs and traceability manifests.
• Source builds require Rust, Node.js/npm, Tauri system prerequisites, and
  FFmpeg/ffprobe available through explicit paths, environment variables, or
  PATH. Packaged desktop builds include their own media tools.

Feedback

Please report reproducible issues with your operating system, installation
artifact, input video format, and the relevant CLI or desktop error output.

FFmpeg dependency ffmpeg-8.1.2-r1

Verified LGPL FFmpeg dependency candidate. It becomes active only after its generated lock PR is merged.