Given a video of a moving object, is it possible to register the 3d gaussian object from the video and optimize the pose of the video? Let's say the video is collected in driving scenario, a random car driven by and is it possible to recover the accurate the pose of that car in each frame? The difference is that the car can only be observed from a constant angle, here we focus more on the poses of the car, and focus less on the accurate reconstruction of that car.
Given a video of a moving object, is it possible to register the 3d gaussian object from the video and optimize the pose of the video? Let's say the video is collected in driving scenario, a random car driven by and is it possible to recover the accurate the pose of that car in each frame? The difference is that the car can only be observed from a constant angle, here we focus more on the poses of the car, and focus less on the accurate reconstruction of that car.