Semantic Segmentation On Pascal Context

评估指标

mIoU

评测结果

各个模型在此基准测试上的表现结果

模型名称	mIoU	Paper Title	Repository
EncNet (ResNet-101)	51.7	Context Encoding for Semantic Segmentation
HRNetV2 HRNetV2-W48	54	Deep High-Resolution Representation Learning for Visual Recognition
CondNet(ResNest-101)	57	CondNet: Conditional Classifier for Scene Segmentation
ViT-Adapter-L (Mask2Former, BEiT pretrain)	68.2	Vision Transformer Adapter for Dense Predictions
SegViT (ours)	65.3	SegViT: Semantic Segmentation with Plain Vision Transformers
ResNeSt-200	58.4	ResNeSt: Split-Attention Networks
OCR (HRNetV2-W48)	56.2	Segmentation Transformer: Object-Contextual Representations for Semantic Segmentation
DGCNet (MS, ResNet-101)	53.7	Dual Graph Convolutional Network for Semantic Segmentation
CAA + Simple decoder (Efficientnet-B7)	60.5	Channelized Axial Attention for Semantic Segmentation -- Considering Channel Relation within Spatial Attention for Semantic Segmentation
LaU-regression-loss (ResNet-101)	53.9	Location-aware Upsampling for Semantic Segmentation
PlainSeg (EVA-02-L)	71.0	Minimalist and High-Performance Semantic Segmentation with Plain Vision Transformers
CAA + CAR (ConvNeXt-Large + JPU)	64.1	CAR: Class-aware Regularizations for Semantic Segmentation
CFM	34.4	Convolutional Feature Masking for Joint Object and Stuff Segmentation
DeepLabV2	45.7	DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs
InternImage-H	70.3	InternImage: Exploring Large-Scale Vision Foundation Models with Deformable Convolutions
Dilated-FCN2s	42.6	Efficient Yet Deep Convolutional Neural Networks for Semantic Segmentation
HRNetV2 + OCR + RMI (PaddleClas pretrained)	59.6	Segmentation Transformer: Object-Contextual Representations for Semantic Segmentation
Asymmetric ALNN	52.8	Asymmetric Non-local Neural Networks for Semantic Segmentation
SenFormer (ResNet-101)	56.6	Efficient Self-Ensemble for Semantic Segmentation
VeryDeep	44.5	Bridging Category-level and Instance-level Semantic Image Segmentation	-

0 of 66 row(s) selected.