The image on the left was taken by a camera. The movie on the right shows the

recovered 3D scene by our model. The red box represents the position of the camera.