Overview of Spatial-MLLM. Our model is composed of a 2D visual encoder, a spatial encoder which is initialized from a feed-forward visual geometry foundation model, a connector, and a large language ...
Some results have been hidden because they may be inaccessible to you
