Welcome to Ovoxador, a cutting-edge system for training athletes like never before. Our solution leverages the power of computer vision and machine learning to process video recordings of team sport games, providing coaches and athletes with unprecedented insights into their performance. Using a combination of neural network models, our solution can detect people and sports equipment on video frames, track players, reidentify them on various tracks, get their texture, determine their pose, and acquire the field texture. These features are then presented in virtual and augmented reality, giving coaches and athletes a 360-degree view of the game and enabling them to analyze their performance in real-time. Whether you are a professional athlete striving for excellence or a coach looking to improve your team's performance, Ovoxador can help you achieve your goals. We invite you to explore our system and discover the many ways it can transform the way you train and play.
Ovoxador harnesses the power of several state-of-the-art neural network models to tackle various tasks and extract valuable insights from video recordings of team sport games. These models include:
- YOLOv8 - A powerful detector that can identify people and sports equipment on video frames with high accuracy.
- DeepSort - A robust tracker that can follow players on video frames and maintain their identities across different camera angles.
- TorchReId - A cutting-edge reidentification model that can recognize players across various tracks and ensure accurate tracking.
- DensePose - A sophisticated model that can extract the texture of players on video frames.
- PARE - A model that can determine the pose of players on video frames, allowing for detailed analysis of their body language and postures.
- LaMa - An advanced image processing model that can acquire field texture.
- Reidentification optimization (now we are using SQLite database to store features)
- Beautiful animated gif for README ;)
- Multithreading support to speed up the work of modules
- Calculation of performance metrics of athletes
- GUI refinement
- VR/AR solution using Unity (will be moved to another repository)
💻 Testing on macOS 💻