How would you approach optimizing an existing computer vision application for speed and accuracy at Thales?
