We present Open3DSG the first approach for learning to predict open-vocabulary 3D scene graphs from 3D point clouds. The advantage of our method is that it can be queried and prompted for any instance ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results