The document summarizes research on characterizing crowds in Piazza Duca d’Aosta in Milan, Italy using computer vision. A YOLO v3 model was used to detect and track pedestrians from camera footage. Trajectories were estimated and filtered to remove errors. Speed, direction, and density heatmaps were generated from the data at 15-minute intervals. Three main clusters of pedestrians were observed near subway entrances and in the center of the square. Speed heatmaps showed higher speeds in the center and uniform speeds near entrances, while density heatmaps illustrated pedestrian concentration patterns over time.