sachinsuresh10/Tunisian_Football_League

Project automates the detection and tracking of football players, referees, and the ball in match videos. It features two phases: object detection and tracking using YOLO and ByteTrack, followed by action recognition with SuperAnnotate. The goal is to enhance analytics in African football leagues, offering valuable insights for decision-making. # Tunisian Football League ## Overview This project aims to develop a **hybrid data collection platform** for automating the detection and tracking of football players, referees, and the ball in match videos. The platform leverages **computer vision models** such as YOLO and ByteTrack to provide real-time insights into match dynamics, focusing on African football leagues where resource constraints limit the adoption of advanced analytics. The project is structured in two phases: 1. **Phase 1**: Object Detection and Tracking 2. **Phase 2**: Action Recognition (Completed Annotation) The goal is to democratize football analytics and bridge the gap between African and European leagues by providing data-driven insights for improved decision-making. --- ## Phase 1: Object Detection and Tracking In the first phase, we implemented **YOLO** (You Only Look Once) for object detection and **ByteTrack** for tracking key entities such as players, referees, and the ball. ### Features of Phase 1: - **Object Detection**: Accurate detection of players, referees, and the ball using YOLOv3, YOLOv5, and YOLOv8 models. - **Object Tracking**: Continuous tracking of objects across frames using ByteTrack, even in cases of occlusion or fast movements. - **Real-Time Insights**: Provides data on player movements, ball possession, and other metrics. ### How to Use: #### 1. **Download the Dataset** The dataset, which has been annotated for football matches, can be accessed via **Roboflow**. - Go to the Powerfoot Computer Vision Project on Roboflow to download different versions of the YOLO model (YOLOv3, YOLOv5, and YOLOv8) along with the corresponding dataset. - You can choose from different dataset formats compatible with various YOLO versions (e.g., COCO, Pascal VOC, YOLO format). #### 2. **Prepare the Input Videos** You will need input football match videos for object detection and tracking. You can use your own videos or download a sample from the provided link. - Download a …