SAVANT - Semi-automated video annotation toolkit

A toolkit for 2D-annotation of video with rotated bounding boxes and object tagging.

• markit - automated annotation (supports YOLO for object detection and classification optical flow for object detection, and AruCo detection for geotags)
• edit - manual annotation or corrections of pre-annotated video, and annotation quality assurance
• trainit - managing datasets and training object detection models

The initial use-case is aerial video annotation of traffic, supported by the use of ASAM OpenLabel output and a traffic-focused default ontology.

Note

This open source project is maintained by RISE Research Institutes of Sweden. See LICENSE file for open source license information.

Installation

Requires Python 3.10+ and uv.

git clone git@github.com:RI-SE/SAVANT.git
cd SAVANT
uv sync

Repository Structure

Directory	Description
markit/	Automated video annotation (README)
edit/	Desktop application for manual label editing and quality assurance (README)
trainit/	YOLO training and dataset tools (README)
utils/	CLI utilities (README)
ontology/	SAVANT ontology definition (README)
schema/	Supported ASAM OpenLabel subset JSON schema (README)

Example Workflow

1. Capture - Record aerial video of traffic scenario
2. Auto-annotate - Run markit to detect and track objects
3. Review - Use edit to correct annotations and verify quality
4. Train - Use trainit to fine-tune models with corrected data
5. Iterate - Re-run markit with improved model

License

SAVANT is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0).

Acknowledgement

This package is developed as part of the SYNERGIES project.

Funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or European Climate, Infrastructure and Environment Executive Agency (CINEA). Neither the European Union nor the granting authority can be held responsible for them.