This repository contains tools for manual data annotation by operators.

It contains the following tools:

• Sports OCR Annotating Tool
• Sports Court Mapping Tool
• Sports Tracks Editing Tool

1. Clone the repository:

git clone https://github.com/darkAlert/sports-anno-tools.git
cd sports-anno-tools

2. Install the virtualenv package and create a new virtual environment variable named sportsvirt and then activate it:

pip3 install virtualenv
python3 -m venv sportsvirt
source sportsvirt/bin/activate

3. Install requirements:

pip3 install -r requirements.txt

The tool is intended for manual editing of predicted players tracks.

Run the tool:

python3 run_tracking_tool.py --data_dir=/path/to/your/videos

The directory /path/to/your/videos has to contain the video clips to be annotated.

After data annotation is complete, the results will be saved to /path/to/your/videos/edited_tracks.json.

If the directory /path/to/your/videos already contains edited_tracks.json, then during the running you will be prompted to load the results from it.

The following hotkeys can be used:

• [ - go to the previous clip
• ] - go to the next clip
• Space - go to the next frame of the clip
• Backspace - go to the previous frame of the clip
• Enter - autoplay / pause
• t - mark the frame as "True Positive" (by default)
• f - mark the frame as "False Positive"
• s - save the results to output json
• Left Mouse Click on the frame area - mark the new position of the player
• Right Mouse Click - delete the marked player position
• Esc - quit (you will be prompted to save the results).