HyperAI超神経

Semantic Segmentation On S3Dis

評価指標

Mean IoU
Number of params
oAcc

評価結果

このベンチマークにおける各モデルのパフォーマンス結果

モデル名
Mean IoU
Number of params
oAcc
Paper TitleRepository
A-CNN62.9N/A87.3A-CNN: Annularly Convolutional Neural Networks on Point Clouds
DeepViewAgg74.741.2M90.1Learning Multi-View Aggregation In the Wild for Large-Scale 3D Semantic Segmentation
RandLA-Net-1.2M87.1RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds
MinkowskiNet65.437.9M-4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks
MuGNet69.8N/A88.5MuGNet: Multi-Resolution Graph Neural Network for Large-Scale Pointcloud Segmentation
EQ-Net77.5N/A-A Unified Query-based Paradigm for Point Cloud Understanding
Feature Geometric Net (FG-Net)70.8N/A88.2FG-Net: Fast Large-Scale LiDAR Point Clouds Understanding Network Leveraging Correlated Feature Mining and Geometric-Aware Modelling
Swin3D-L79.8N/A92.4Swin3D: A Pretrained Transformer Backbone for 3D Indoor Scene Understanding
JSENet67.7N/A-JSENet: Joint Semantic Segmentation and Edge Detection Network for 3D Point Clouds
BIM-Net---Fully Automated Scan-to-BIM Via Point Cloud Instance Segmentation
PointCNN65.4N/A88.1A-CNN: Annularly Convolutional Neural Networks on Point Clouds
PointASNL68.7N/A88.8PointASNL: Robust Point Clouds Processing using Nonlocal Neural Networks with Adaptive Sampling
KPConv70.614.1M-KPConv: Flexible and Deformable Convolution for Point Clouds
PointCNN65.4N/A-Point Transformer
BAAF-Net72.2N/A88.9Semantic Segmentation for Real Point Cloud Scenes via Bilateral Augmentation and Adaptive Fusion
PPT + SparseUNet78.1N/A92.2Towards Large-scale 3D Representation Learning with Multi-dataset Point Prompt Training
JSNet61.7N/A88.7JSNet: Joint Instance and Semantic Segmentation of 3D Point Clouds
ShellNet66.8N/A-ShellNet: Efficient Point Cloud Convolutional Neural Networks using Concentric Shells Statistics-
SPGraph62.1N/A85.5A-CNN: Annularly Convolutional Neural Networks on Point Clouds
3P-RNN56.3N/A86.9A-CNN: Annularly Convolutional Neural Networks on Point Clouds
0 of 54 row(s) selected.