PointGroup: Dual-Set Point Grouping for 3D Instance Segmentation
This work addresses the problem of 3D instance segmentation for scene understanding, offering a novel method that significantly improves accuracy over existing solutions, though it is incremental in advancing grouping techniques.
The paper tackles 3D instance segmentation for point clouds by proposing PointGroup, a bottom-up architecture that groups points using dual sets of coordinates, achieving state-of-the-art performance with 63.6% mAP on ScanNet v2 and 64.0% on S3DIS, compared to previous bests of 54.9% and 54.4%.
Instance segmentation is an important task for scene understanding. Compared to the fully-developed 2D, 3D instance segmentation for point clouds have much room to improve. In this paper, we present PointGroup, a new end-to-end bottom-up architecture, specifically focused on better grouping the points by exploring the void space between objects. We design a two-branch network to extract point features and predict semantic labels and offsets, for shifting each point towards its respective instance centroid. A clustering component is followed to utilize both the original and offset-shifted point coordinate sets, taking advantage of their complementary strength. Further, we formulate the ScoreNet to evaluate the candidate instances, followed by the Non-Maximum Suppression (NMS) to remove duplicates. We conduct extensive experiments on two challenging datasets, ScanNet v2 and S3DIS, on which our method achieves the highest performance, 63.6% and 64.0%, compared to 54.9% and 54.4% achieved by former best solutions in terms of mAP with IoU threshold 0.5.