Player Classification Tool (UPF)

Our player classification tool transforms single-camera soccer footage into actionable insights. It identifies all five key participant roles in a match—referee, outfield players from both teams, and goalkeepers—by analyzing scene segmentations. Using advanced deep learning techniques, including a convolutional network and a transformer encoder, the tool processes player data to accurately determine their roles. GitHub

Action Spotting Tool (UPF)

Our action spotting tool uses a transformer-based deep learning architecture to analyze soccer match videos with remarkable precision. By combining visual and audio data, it captures both short- and long-term contexts around each event. The system processes video frames to extract features, identifies patterns using a transformer encoder, and classifies actions with high accuracy. A final post-processing step ensures reliable predictions, making it a powerful tool for understanding match dynamics. GitHub