We pre-trained Atlas from scratch to take multimodal inputs, including camera movement, and turn it into 3D grounded views.
Atlas puts you in control: direct the views, reconstruct real spaces with your inputs, and build explorable worlds.
Read more:
顯示更多