Abstract: Highlights•This paper proposes a holistic 3D plane recovery method that directly detects plane instances along with their corresponding 3D planar parameters.•This paper proposes the use of a parallel dilated mask head for boundary predictions, which is lightweight yet effectively imposes boundary supervision, resulting in high-quality 3D plane recovery.•Comprehensive experiments show the proposed method achieves state-of-the-art performance on ScanNet and NYUv2 datasets, outperforming rivals in plane segmentation quality and normal vector regression precision, verifying its effectiveness.
Loading