Closed waveleaf27 closed 1 year ago
Hi @waveleaf27 , For training, the full resolution of Waymo sequences are used, which is [1280x1920]; For evaluation, all of our models and baseline models are evaluated with downscale=2, which means [640x960]. This is just for saving time while also ensuring a fair comparison.
Hi, can you provide the image resolution used in the experiment? THX