Abstract: With the growing interest in 3D Gaussian Splatting (3DGS) for scene analysis, current approaches typically extract 2D features from multi-view images via pre-trained models before embedding ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results