EPFL MultiCam-GT v2

Updated version of EPFL MultiCam-GT Tool, a multiple view object annotation tool, now with:

• Elevation model: Improved ground plane representation through mesh-based elevation modeling, for object annotation in varying terrains.
• Enhanced Object Transformation: The new version includes object representation using dimensions and spatial orientation.
• Tracklet Merging: Functionality for merging tracklets, providing more streamlined object tracking.
• Trajectory Visualization: Features new tools for visualizing trajectories, enhancing understanding and analysis of object movement.
• Database Integration: All transformation and tracking data is now stored in the database, enabling more efficient data management and retrieval.

TODO

Usage

Define the DATABASES variable in the settings.py file to point to your database. The application is configured to use a PostgreSQL database. You can use the following configuration:

DATABASES = {
    'default': {
        'ENGINE': 'django.db.backends.postgresql_psycopg2',
        'NAME': 'scout',
        'USER': 'scout',
        'PASSWORD': '',
        'HOST': 'localhost',
        'PORT': '',
    }
}

You can load the SCOUT data into the database by running the following command:

python manage.py shell < data_process.py

By default, the worker name will be 'SCOUT'. You can change it by editing the data_process.py file. You can then run the following commands to run the application:

conda env create -n ENVNAME --file ENV.yml
pip install -r requirements.txt
conda install postgresql
initdb -D scout   
chmod -R 700 scout
pg_ctl -D scout -l logfile start
createuser scout
createdb --owner=scout scout

python manage.py migrate
python import_annotations.py --dataset SCOUT --worker workerID --input /cvlabdata2/cvlab/scout/annotations/raw/annotations_sequence_1_raw.json
python manage.py runserver 0.0.0.0:4444

You can now access the application at http://localhost:4444

Backup via:

pg_dump -U scout -h localhost -d scout -F c -f annotation_snapshot_280125.dump

### Acknowledgements
This project was based on the original [MultiCam-GT Tool](https://github.com/cvlab-epfl/multicam-gt) developed by the Computer Vision Laboratory at EPFL.