Home
Desktop annotation platform for Machine Learning datasets
Images · Video · Time Series · Tabular Data
Welcome to the Annotix documentation. Annotix is a cross-platform desktop application for creating, managing, and exporting annotated ML datasets. It pairs a React 19 frontend with a high-performance Rust backend through Tauri 2.
- Getting Started — Installation, prerequisites, first run, and development setup.
- Annotation Tools — The 7 canvas tools (BBox, OBB, Polygon, Mask, Keypoints, Landmarks, Select), zoom, pan, class selection, undo/redo, and image adjustments.
- Video Annotation — Frame extraction, tracks, keyframes, interpolation, bake, and timeline controls.
- Time Series — CSV import, visualization, and the 5 annotation types (point, range, classification, event, anomaly).
- Tabular Data — Data editor, column selection, and classical ML training.
- Integrated ML Training — 19 ML backends, 4 execution modes, presets, real-time metrics, cloud providers, and model export.
- Inference — ONNX and PyTorch model inference, batch processing, NMS, and confidence thresholds.
- Export and Import — 11 export formats, 8 import formats with auto-detection, coordinate systems, and file structures.
- P2P Collaboration — Real-time annotation with Iroh (QUIC), roles, permissions, image locking, work distribution, and CRDT sync.
- Browser Automation — Free Google Colab training and LLM queries without API keys via Chrome DevTools Protocol.
- Keyboard Shortcuts — Full shortcut reference, customization system, and conflict detection.
- Languages — 10 supported languages with lazy loading.
- Architecture — Storage model, IPC layer, state management, handler pattern, and project structure.
| Feature | Details |
|---|---|
| Annotation Tools | 7 tools on high-performance Konva canvas |
| Data Types | Images, Video, Time Series, Tabular |
| ML Backends | 19 (YOLO, RT-DETR, MMDetection, Detectron2, timm, SMP, etc.) |
| Execution Modes | Local, Download Package, Cloud, Browser Automation |
| Export Formats | 11 with real-time progress |
| Import Formats | 8 with automatic format detection |
| P2P Collaboration | Iroh QUIC, no central server |
| Languages | 10 (de, en, es, fr, it, ja, ko, pt, ru, zh) |
| Platforms | Windows, macOS, Linux |
| Tech Stack | React 19 + Rust + Tauri 2 |
| Version | 2.3.1 |