Semantic Segmentation On S3Dis Area5

Metrics

Number of params

mAcc

mIoU

oAcc

Results

Performance results of various models on this benchmark

					Paper Title
Sonata + PTv3	-	81.6	76.0	93.0	Sonata: Self-Supervised Learning of Reliable Point Representations
OmniVec	-	-	75.9	-	OmniVec: Learning robust representations with cross modal sharing
PTv3 + PPT	-	80.1	74.7	92.0	Point Transformer V3: Simpler, Faster, Stronger
Swin3D-L	N/A	80.5	74.5	92.7	Swin3D: A Pretrained Transformer Backbone for 3D Indoor Scene Understanding
DITR	-	-	74.1	-	DINO in the Room: Leveraging 2D Foundation Models for 3D Segmentation
DeLA	7.0M	80.0	74.1	92.2	Decoupled Local Aggregation for Point Cloud Learning
Ours	-	80.2	73.6	93.0	Beyond local patches: Preserving global–local interactions by enhancing self-attention via 3D point cloud tokenization
Pamba	-	-	73.5	-	Pamba: Enhancing Global Interaction in Point Clouds via State Space Model
ConDaFormer	-	78.9	73.5	92.4	ConDaFormer: Disassembled Transformer with Local Structure Enhancement for 3D Point Cloud Understanding
LPFP(Point Transformer*)	31.2M	78.7	73.5	92.0	A Large-Scale Network Construction and Lightweighting Method for Point Cloud Semantic Segmentation
KPConvX-L	-	78.7	73.5	91.7	KPConvX: Modernizing Kernel Point Convolution with Kernel Attention
SPG(PTv2)	-	79.5	73.3	91.9	Subspace Prototype Guidance for Mitigating Class Imbalance in Point Cloud Semantic Segmentation
PointHR	-	78.7	73.2	91.8	PointHR: Exploring High-Resolution Architectures for 3D Point Cloud Segmentation
PonderV2 + SparseUNet	-	79.0	73.2	92.2	PonderV2: Pave the Way for 3D Foundation Model with A Universal Pre-training Paradigm
PPT + SparseUNet	N/A	78.2	72.7	91.5	Towards Large-scale 3D Representation Learning with Multi-dataset Point Prompt Training
PTv2	N/A	78.0	72.6	91.6	Point Transformer V2: Grouped Vector Attention and Partition-based Pooling
PT + ERDA	-	-	72.6	-	-
SAT (FAT)	N/A	78.8	72.6	-	SAT: Size-Aware Transformer for 3D Point Cloud Semantic Segmentation
PointVector-XL	-	78.1	72.3	91	PointVector: A Vector Representation In Point Cloud Analysis
WindowNorm+StratifiedTransformer	N/A	78.2	72.2	91.4	Window Normalization: Enhancing Point Cloud Understanding by Unifying Inconsistent Point Densities

0 of 60 row(s) selected.

Command Palette

Semantic Segmentation On S3Dis Area5

Metrics

Results