S3DIS

Overview edit

Variants: S3DIS, S3DIS Area5

Associated Benchmarks

This dataset is used in 6 benchmarks:

• Semantic Segmentation - Metrics: Mean IoU, mAcc, oAcc, FLOPs, Number of params, mIoU, Params (M)
• 3D Semantic Segmentation - Metrics: mIoU (Area-5), mIoU (6-Fold), mAcc
• 3D Object Detection - Metrics: [email protected], [email protected]
• 3D Open-Vocabulary Instance Segmentation - Metrics: AP50 Base B8/N4 , AP50 Novel B8/N4, AP50 Base B6/N6, AP50 Novel B6/N6
• Panoptic Segmentation - Metrics: PQ, RQ, SQ, PQ (with stuff), RQ (with stuff), SQ (with stuff), Params (M)
• 3D Instance Segmentation - Metrics: AP@50, mAP, mPrec, mRec, mIoU, mAcc, mCov, mWCov

Recent Benchmark Submissions

Research Papers

Recent papers with results on this dataset:

• Sonata: Self-Supervised Learning of Reliable Point Representations (2025) -
• MSTA3D: Multi-scale Twin-attention for 3D Instance Segmentation (2024) -
• UniDet3D: Multi-dataset Indoor 3D Object Detection (2024) -
• Scalable 3D Panoptic Segmentation As Superpoint Graph Clustering (2024) -
• SPGroup3D: Superpoint Grouping Network for Indoor 3D Object Detection (2023) -
• Open3DIS: Open-Vocabulary 3D Instance Segmentation with 2D Mask Guidance (2023) -
• Point Transformer V3: Simpler, Faster, Stronger (2023) -
• OneFormer3D: One Transformer for Unified Point Cloud Segmentation (2023) -
• PonderV2: Pave the Way for 3D Foundation Model with A Universal Pre-training Paradigm (2023) -
• OpenIns3D: Snap and Lookup for 3D Open-vocabulary Instance Segmentation (2023) -
• Towards Large-scale 3D Representation Learning with Multi-dataset Point Prompt Training (2023) -
• Lowis3D: Language-Driven Open-World Instance-Level 3D Scene Understanding (2023) -
• Efficient 3D Semantic Segmentation with Superpoint Transformer (2023) -
• Point-GCC: Universal Self-supervised 3D Scene Pre-training via Geometry-Color Contrast (2023) -
• Swin3D: A Pretrained Transformer Backbone for 3D Indoor Scene Understanding (2023) -
• ISBNet: a 3D Point Cloud Instance Segmentation Network with Instance-aware Sampling and Box-aware Dynamic Convolution (2023) -
• Top-Down Beats Bottom-Up in 3D Instance Segmentation (2023) -
• TR3D: Towards Real-Time Indoor 3D Object Detection (2023) -
• Window Normalization: Enhancing Point Cloud Understanding by Unifying Inconsistent Point Densities (2022) -
• PLA: Language-Driven Open-Vocabulary 3D Scene Understanding (2022) -