π· IOD2VOT π½οΈ Turn any Image object detection model into a video object tracking model. Why try to train a complicated video model when a simple image model + intuiative huristics will do fine?