kabaddi_AI_analytics

This repository provides tools for analyzing Kabaddi videos using AI-based techniques. It includes functionalities for team classification and overall analytics derived from the input footage.

Getting Started

Prerequisites:

• Required Python packages (install via requirements.txt)
• Input video files in data/videos/ directory

Running the Tool: Use the following command to process a video and perform team classification:

python main.py \
    --source_video_path data/videos/kabedi.mp4 \
    --target_video_path data/videos/results/kabadi_teamclass.mp4 \
    --device cpu \
    --mode TEAM_CLASSIFICATION

Example of Classification

Model Files

The football .pt model file is sourced from the SportVision repository. The team classification logic and main analytical approach are adapted from the Roboflow Sport repository https://github.com/roboflow/sports.

And the kabbadi .pt models are fine-tuned on kabbadi dataset.

2D Map of the raider

Time spend on the opponent court