Workflows¶
Standard Video-First Workflow¶
Annolid supports direct video labeling without mandatory frame extraction.
- Prepare your video data.
- Open video in Annolid.
- Jump to a representative frame.
- Label instances on that frame.
- Run tracking/propagation.
- Review and correct predictions.
- Export annotations or downstream training data.
Practical Notes¶
- Use extraction tools only when you explicitly need frame-level datasets.
- Keep annotations incremental and review high-motion sections early.
- Prefer short review loops over long unattended tracking runs.
Advanced References¶
- Book walkthrough: book/content/README.md