Papers › Point Transformer V2: Grouped Vector Attention and Partition-based Pooling

Point Transformer V2: Grouped Vector Attention and Partition-based Pooling

11 Oct 2022arXiv:2210.05666archive 2025-07-28

Xiaoyang Wu, Yixing Lao, Li Jiang, Xihui Liu, Hengshuang Zhao

As a pioneering work exploring transformer architecture for 3D point cloud understanding, Point Transformer achieves impressive results on multiple highly competitive benchmarks. In this work, we analyze the limitations of the Point Transformer and propose our powerful and efficient Point Transformer V2 model with novel designs that overcome the limitations of previous work. In particular, we first propose group vector attention, which is more effective than the previous version of vector attention. Inheriting the advantages of both learnable weight encoding and multi-head attention, we present a highly effective implementation of grouped vector attention with a novel grouped weight encoding layer. We also strengthen the position information for attention by an additional position encoding multiplier. Furthermore, we design novel and lightweight partition-based pooling methods which enable better spatial alignment and more efficient sampling. Extensive experiments show that our model achieves better performance than its predecessor and achieves state-of-the-art on several challenging 3D point cloud understanding benchmarks, including 3D point cloud segmentation on ScanNet v2 and S3DIS and 3D point cloud classification on ModelNet40. Our code will be available at https://github.com/Gofinge/PointTransformerV2.

PaperPDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

Pointcept/PointTransformerV2 officialmentioned on GitHubpytorch report
Pointcept/Pointcept officialmentioned on GitHubpytorchMIT report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

3D Point Cloud Classification3D Semantic SegmentationLIDAR Semantic SegmentationPoint Cloud ClassificationPoint Cloud SegmentationSemantic Segmentation

2 archive task tags without a task page not shown.

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
3D Point Cloud Classification ModelNet40 PTv2 Mean Accuracy 91.6 #18 of 111 Archive leaderboard report
3D Point Cloud Classification ModelNet40 PTv2 Overall Accuracy 94.2 #18 of 111 Archive leaderboard report
3D Semantic Segmentation S3DIS PointTransformerV2 mIoU (Area-5) 71.6 #2 of 6 Archive leaderboard report
3D Semantic Segmentation ScanNet++ PTv2 Top-1 IoU 0.445 #8 of 8 Archive leaderboard report
3D Semantic Segmentation ScanNet++ PTv2 Top-3 IoU 0.688 #8 of 8 Archive leaderboard report
3D Semantic Segmentation SemanticKITTI PTv2 test mIoU 72.6% #9 of 45 Archive leaderboard report
3D Semantic Segmentation SemanticKITTI PTv2 val mIoU 70.3% #9 of 45 Archive leaderboard report
3D Semantic Segmentation nuScenes PTv2 mIoU 82.6% #1 of 3 Archive leaderboard report
LIDAR Semantic Segmentation nuScenes PTv2 test mIoU 0.826 #3 of 36 Archive leaderboard report
LIDAR Semantic Segmentation nuScenes PTv2 val mIoU 0.802 #3 of 36 Archive leaderboard report
Semantic Segmentation S3DIS Area5 PTv2 Number of params N/A #18 of 61 Archive leaderboard report
Semantic Segmentation S3DIS Area5 PTv2 mAcc 78.0 #18 of 61 Archive leaderboard report
Semantic Segmentation S3DIS Area5 PTv2 mIoU 72.6 #18 of 61 Archive leaderboard report
Semantic Segmentation S3DIS Area5 PTv2 oAcc 91.6 #18 of 61 Archive leaderboard report
Semantic Segmentation ScanNet PTv2 test mIoU 75.2 #21 of 45 Archive leaderboard report
Semantic Segmentation ScanNet PTv2 val mIoU 75.4 #21 of 45 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

AdamAttentionBPEDense ConnectionsDropoutLabel SmoothingLayer NormalizationLinear LayerMulti-Head AttentionPosition-Wise Feed-Forward LayerResidual ConnectionSoftmaxTransformer

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections