- JavaScript 72.6%
- Python 19.9%
- CSS 6.8%
- HTML 0.3%
- PowerShell 0.2%
- Autre 0.2%
| assets | ||
| client | ||
| experiments | ||
| .gitignore | ||
| build.js | ||
| install.ps1 | ||
| install.sh | ||
| LICENSE.txt | ||
| package-lock.json | ||
| package.json | ||
| README.md | ||
| requirements.txt | ||
How to run
Install Node.js 18 or more recent. You can use NVM to manage recent Node versions on older Linux distributions.
Once Node is installed, install the dependencies with:
npm install
Create and activate a Conda environment for Python dependencies :
conda create -n trakking python=3.10
conda activate trakking
Install Python dependencies :
chmod +x install.sh
./install.sh
You can then run the web app with auto-rebuild after each change with:
npm run develop
The web app is available at http://localhost:8000/
To start a PyTracker, use the following command from the root directory of the project :
python client/backend/run_server.py [-h] --model_path MODEL_PATH [--host HOST]
The port is automatically assigned to a specific PyTracker model, and the host defaults to localhost (127.0.0.1).
Currently available models :
python client/backend/run_server.py --model_path itracker_baseline.tar
python client/backend/run_server.py --model_path itracker_mpiiface.tar
Note: If you select a tracker from the tracker list in the interface, make sure to start its corresponding server beforehand using the appropriate command above.
Tips
If you add a tracker (in client/trackers/) or an experiment (in experiments/) you need to stop and restart the development script.
Use Alt + O to show the tracker prediction in the UI.