UnScene3D: Unsupervised 3D Instance Segmentation for Indoor Scenes

arXiv:2303.14541v2 Announce Type: replace
Abstract: 3D instance segmentation is fundamental to geometric understanding of the world around us. Existing methods for instance segmentation of 3D scenes rely on supervision from expensive, manual 3D annotations. We propose UnScene3D, the first fully unsupervised 3D learning approach for class-agnostic 3D instance segmentation of indoor scans. UnScene3D first generates pseudo masks by leveraging self-supervised color and geometry features to find potential object regions. We operate on a basis of geometric oversegmentation, enabling efficient representation …

3d scenes abstract annotations arxiv class cs.cv fundamental instance scans segmentation supervision type understanding unsupervised world