Abstract: Recent open-world representation learning approaches have leveraged CLIP to enable zero-shot 3D object recognition. However, performance on real point clouds with occlusions still falls ...
Stop coding without these extensions ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results