Papers › Beyond local patches: Preserving global–local interactions by enhancing self-attention...

Beyond local patches: Preserving global–local interactions by enhancing self-attention via 3D point cloud tokenization

1 Nov 2024Pattern Recognition, Vol. 155 110712 2024 11archive 2025-07-28

M.Q. Khan, M. Shahzad, S.A. Khan, M.M. Fraz, X.X. Zhu

Transformer-based architectures have recently shown impressive performance on various point cloud understanding tasks such as 3D object shape classification and semantic segmentation. Particularly, this can be attributed to their self-attention mechanism, which has the ability to capture long-range dependencies. However, current methods have constrained it to operate in local patches due to its quadratic memory constraints. This hinders their generalization ability and scaling capacity due to the loss of non-locality in early layers. To tackle this issue, we propose a window-based transformer architecture that captures long-range dependencies while aggregating information in the local patches. We do this by interacting each window with a set of global point cloud tokens — a representative subset of the entire scene — and augmenting the local geometry through a 3D Histogram of Oriented Gradients (HOG) descriptor. Through a series of experiments on segmentation and classification tasks, we show that our model exceeds the state-of-the-art on S3DIS semantic segmentation (+1.67% mIoU), ShapeNetPart part segmentation (+1.03% instance mIoU) and performs competitively on ScanObjectNN 3D object classification. he code and trained models shall be made publicly available.

PaperPDF

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

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

Tasks

3D Classification3D Object Classification3D Part Segmentation3D Point Cloud Classification3D Semantic SegmentationSegmentationSemantic Segmentation

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
3D Part Segmentation ShapeNet-Part Ours Instance Average IoU 88.1 #2 of 67 Archive leaderboard report
3D Point Cloud Classification ScanObjectNN Ours Mean Accuracy 87.2 #30 of 77 Archive leaderboard report
3D Point Cloud Classification ScanObjectNN Ours Overall Accuracy 89.0 #30 of 77 Archive leaderboard report
Semantic Segmentation S3DIS Area5 Ours mAcc 80.2 #8 of 61 Archive leaderboard report
Semantic Segmentation S3DIS Area5 Ours mIoU 73.6 #8 of 61 Archive leaderboard report
Semantic Segmentation S3DIS Area5 Ours oAcc 93.0 #8 of 61 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

SET

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