AI Paper Atlas

Research map / Detection and segmentation / 3D detection and driving

Superpoint Transformer for 3D Scene Instance Segmentation

Jiahao Sun, Chunmei Qing, Junpeng Tan, Xiangmin Xu

AAAI 2023 · Detection and segmentation › 3D detection and driving · approach: 3D objectness and scene segmentation · 125 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use SPFormer's superpoint-query instance segmentation instead of the strongest listed baselineScanNetv2 hidden test · ScanNetv2 hidden testmAP: 50.6 -> 54.9Improved
Use SPFormer rather than SSTNetScanNetv2 validation · ScanNetv2 validationmAP: 49.4 -> 56.3Improved
Use SPFormer rather than the strongest listed baselineS3DIS Area 5 · S3DIS Area 5AP50: 66.1 -> 66.8Improved
Use SPFormer rather than the strongest listed baselineS3DIS 6-fold cross-validation · S3DIS 6-fold cross-validationAP50: 68.9 -> 69.2Improved
Add superpoint pooling between the point backbone and query decoder (ablation)SPFormer · ScanNetv2 validationmAP: 34.3 -> 55.0Improved
Replace box matching with mask matching (ablation)SPFormer with superpoint pooling · ScanNetv2 validationmAP: 49.9 -> 55.0Improved
Add the IoU-aware score branch (ablation)SPFormer with superpoint pooling and mask matching · ScanNetv2 validationmAP: 55.0 -> 56.3Improved
Add iterative predictions from each decoder layer (ablation)SPFormer transformer decoder · ScanNetv2 validationmAP: 51.0 -> 52.5Improved
Add superpoint attention masks (ablation)SPFormer with iterative prediction · ScanNetv2 validationmAP: 52.5 -> 56.0Improved
Add position encoding to query vectors (ablation)SPFormer with iterative prediction and attention masks · ScanNetv2 validationmAP: 56.0 -> 55.6Hurt
Place cross-attention before self-attention (ablation)SPFormer transformer decoder · ScanNetv2 validationmAP: 56.0 -> 56.3Improved
Use 6 rather than 1, 3, or 12 transformer decoder layers (ablation)SPFormer · ScanNetv2 validationmAP: 49.1; 54.7; 56.3; 55.3Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy