The client's primary goal for the project was to replicate the 3D pose estimation technology used in professional football leagues' VAR systems. The focus was on implementing the state of the art deep learning model for accurate player detection and pose estimation using multiple cameras, to explore the potential of Computer Vision.

● As the developer of the project, my role involved several key tasks. Firstly, I researched and implemented the YOLOv7 model for object detection, with a specific focus on football player recognition. I fine-tuned the model to ensure high accuracy in detecting players across various camera angles and lighting conditions.

● Additionally, I integrated the 3D pose estimation module into the system, leveraging the information obtained from multiple cameras. This involved aligning the detected player positions to reconstruct a 3D representation of their poses, providing a more comprehensive analysis of player movements during a match.

● I also worked on optimizing the system for real-time performance, considering the dynamic nature of football games. This required efficient processing of video feeds from multiple cameras and the seamless integration of the 3D pose estimation results.

The project was a success, achieving the client's goal of implementing 3D pose estimation for football players using YOLOv7. The system demonstrated accurate player detection and provided valuable insights into player movements in three-dimensional space. This not only addressed offside calls but also opened up possibilities for in-depth match analysis, player performance evaluation, and tactical insights.