This sinapsis package provides a robust implementation for multi-object tracking with the Facebook Research's Co-Tracker model. It includes:

• Templates for multi-object tracking using Co-Tracker, offering flexible offline, online, and visualization modes.
• Efficient processing and visualization of tracking results directly on video frames for clear output.
• Tools for handling dynamic tracking across frames, including padding, line width, and trace settings.

For specific instructions and further details, see the README.md.

This Sinapsis package provides a comprehensive solution for object tracking with the ByteTrack algorithm. It includes:

• A template for object tracking using ByteTrack, designed to handle real-time multi-object tracking in videos.
• Detection processing and updates with configurable parameters for track activation, matching, and occlusion handling, improving accuracy and stability.

For more details, see the README.md.

This Sinapsis package provides robust templates for multi-object tracking, leveraging the trackers library. It integrates powerful algorithms like SORT and DeepSORT into the Sinapsis ecosystem.

• SORT Tracker: A simple and efficient motion-based tracker ideal for high-speed applications.
• DeepSORT Tracker: An advanced tracker that enhances SORT by incorporating appearance features using a configurable Re-Identification (Re-ID) model. This makes it more robust against occlusions and helps maintain object identities in complex scenes.
• Flexible Configuration: Offers extensive attributes to fine-tune tracker behavior, including support for various Re-ID models via the timm library.

For specific instructions and further details, see the README.md.

IMPORTANT: This Docker image depends on the sinapsis-nvidia:base image. For detailed instructions, please refer to the Sinapsis README.

1. Build the sinapsis-trackers image:

docker compose -f docker/compose.yaml build

2. Start the container:

For sinapsis-cotracker

docker compose -f docker/compose_tracker.yaml up sinapsis-cotracker-gradio -d

For sinapsis-supervision with default bytetrack-ultralytics demo

docker compose -f docker/compose_tracker.yaml up sinapsis-supervision-gradio -d

For sinapsis-supervision with bytetrack-dfine demo

export DFINE_CONFIG_DOWNLOAD=True
export AGENT_CONFIG_PATH=/app/sinapsis_supervision/configs/bytetrack_dfine_demo.yml
docker compose -f docker/compose_tracker.yaml up sinapsis-supervision-gradio -d

3. Check the status:

For sinapsis-cotracker

docker logs -f sinapsis-cotracker-gradio

For sinapsis-supervision

docker logs -f sinapsis-supervision-gradio

4. The logs will display the URL to access the webapp, e.g.,:

Running on local URL:  http://127.0.0.1:7860

NOTE: The local URL can be different, please check the logs

5. To stop the app:

docker compose -f docker/compose_tracker.yaml down

1. Create the virtual environment and sync the dependencies:

uv sync --frozen --extra cotracker

2. Install the sinapsis-trackers package:

uv pip install sinapsis-trackers[all] --extra-index-url https://pypi.sinapsis.tech

3. Run the webapp:

For demo running default cotracker-online agent config.

uv run webapps/tracking_demo.py

For demo running bytrack-ultralytics agent config.

export AGENT_CONFIG_PATH="packages/sinapsis_supervision/src/sinapsis_supervision/configs/bytetrack_ultralytics_demo.yml"
uv run webapps/tracking_demo.py

For demo running bytetrack-dfine agent config.

export DFINE_CONFIG_DOWNLOAD=True
export AGENT_CONFIG_PATH="packages/sinapsis_supervision/src/sinapsis_supervision/configs/bytetrack_dfine_demo.yml"
uv run webapps/tracking_demo.py

4. The terminal will display the URL to access the webapp, e.g.:

Running on local URL:  http://127.0.0.1:7860

NOTE: The local URL can be different, please check the output of the terminal.