How to run this version of AnyLabelling

1. Open the Perception Docker container

export TAG_NAME="latest" && xhost +local:docker && docker compose -f docker_files/docker-compose.yml run --rm perception`

2. Clone AnyLabelling from our fork

This should be done inside VS Code, so that you automatically use your Git credentials:

cd ~/ws_lab0/src
git clone [email protected]:LAB0-Inc/anylabeling.git

3. set AnyLabeling up

cd anylabeling
python3 -m pip install --user --no-cache-dir --break-system-packages  -r requirements-gpu.txt
sudo apt-get update
sudo apt-get install pyqt5-dev-tools
pyrcc5 -o anylabeling/resources/resources.py anylabeling/resources/resources.qrc

4. Run AnyLabelling with:

	cd ~/ws_lab0/src/anylabeling && python3 -m anylabeling.app

Changes w.r.t. the original code

• Contours of boxes are drawn with 5 random colors, rather than all the same.
• Segments are drawn in semi-transparent white.
• Added shortcuts for hiding [Z] and for visualizing [X] the annotations.
• The default zoom level was adjusted so that the images are fully visibile.
• The images are visualized as raw: without smoothing and without the JPEG artifacts that appear in the original version of the code.
• The cross line is not visualized by default.
• When creating a new annotation, the latest label is automatically applied to it.
• Epsilon (the distance at which the mouse pointer sticks to contours) was reduced to 5 pixels. This allows selecting one small segment (by clicking in the middle of it, away from the contour) without having to first zoom in on it.