1 paper
Mayank Anand, Mohammad Saqlain, Kyan Mahajan +3
Tracking points in videos is typically formulated as frame-to-frame correspondence, where each point is matched locally to the next frame. While this works over short horizons, err…