| Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation | ✓ Link | 89.0% | | | DeepLabv3+ (Xception-65-JFT) | 2018-02-07 |
| Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation | ✓ Link | 89.0% | | | DeepLabv3+ (Xception-JFT) | 2018-02-07 |
| Rethinking Atrous Convolution for Semantic Image Segmentation | ✓ Link | 86.9% | | | DeepLabv3-JFT | 2017-06-17 |
| Stacked Deconvolutional Network for Semantic Segmentation | | 86.6% | | | CASIA_IVA_SDN | 2017-08-16 |
| Learning a Discriminative Feature Network for Semantic Segmentation | ✓ Link | 86.2% | | | Smooth Network with Channel Attention Block | 2018-04-25 |
| Squeeze-and-Attention Networks for Semantic Segmentation | ✓ Link | 86.1% | | | SANet (pretraining on COCO dataset) | 2019-09-08 |
| Context Encoding for Semantic Segmentation | ✓ Link | 85.9% | | | EncNet | 2018-03-23 |
| Is Attention Better Than Matrix Decomposition? | ✓ Link | 85.9% | | | HamNet w/o COCO (ResNet-101) | 2021-09-09 |
| Auto-DeepLab: Hierarchical Neural Architecture Search for Semantic Image Segmentation | ✓ Link | 85.6% | | | Auto-DeepLab-L | 2019-01-10 |
| Pyramid Scene Parsing Network | ✓ Link | 85.4% | | | PSPNet | 2016-12-04 |
| Wider or Deeper: Revisiting the ResNet Model for Visual Recognition | ✓ Link | 84.9% | | | ResNet-38 MS COCO | 2016-11-30 |
| Segmentation Transformer: Object-Contextual Representations for Semantic Segmentation | ✓ Link | 84.5% | | | OCR (HRNetV2-W48) | 2019-09-24 |
| Segmentation Transformer: Object-Contextual Representations for Semantic Segmentation | ✓ Link | 84.3% | | | OCR (ResNet-101) | 2019-09-24 |
| RefineNet: Multi-Path Refinement Networks for High-Resolution Semantic Segmentation | ✓ Link | 84.2% | | | Multipath-RefineNet | 2016-11-20 |
| ShelfNet for Fast Semantic Segmentation | ✓ Link | 84.2% | | | ShelfNet | 2018-11-27 |
| Beyond Self-attention: External Attention using Two Linear Layers for Visual Tasks | ✓ Link | 84% | | | EANet (ResNet-101) | 2021-05-05 |
| Large Kernel Matters -- Improve Semantic Segmentation by Global Convolutional Network | ✓ Link | 83.6% | | | Large Kernel Matters | 2017-03-08 |
| Triply Supervised Decoder Networks for Joint Detection and Segmentation | | 83.3% | | | TripleNet | 2018-09-25 |
| Squeeze-and-Attention Networks for Semantic Segmentation | ✓ Link | 83.2% | | | SANet | 2019-09-08 |
| Understanding Convolution for Semantic Segmentation | ✓ Link | 83.1% | | | TuSimple | 2017-02-27 |
| Context Encoding for Semantic Segmentation | ✓ Link | 82.9% | | | EncNet (ResNet-101) | 2018-03-23 |
| Learning a Discriminative Feature Network for Semantic Segmentation | ✓ Link | 82.7% | | | DFN (ResNet-101) | 2018-04-25 |
| Not All Pixels Are Equal: Difficulty-aware Semantic Segmentation via Deep Layer Cascade | ✓ Link | 82.7% | | | Deep Layer Cascade (LC) | 2017-04-05 |
| Light-Weight RefineNet for Real-Time Semantic Segmentation | ✓ Link | 82.7% | | | Light-Weight-RefineNet-152 | 2018-10-08 |
| Dual Attention Network for Scene Segmentation | ✓ Link | 82.6% | | | DANet (ResNet-101) | 2018-09-09 |
| Pyramid Scene Parsing Network | ✓ Link | 82.6% | | | PSPNet (ResNet-101) | 2016-12-04 |
| Light-Weight RefineNet for Real-Time Semantic Segmentation | ✓ Link | 82.0% | | | Light-Weight-RefineNet-101 | 2018-10-08 |
| Light-Weight RefineNet for Real-Time Semantic Segmentation | ✓ Link | 81.1% | | | Light-Weight-RefineNet-50 | 2018-10-08 |
| Fast, Exact and Multi-Scale Inference for Semantic Image Segmentation with Deep Gaussian CRFs | ✓ Link | 80.2% | | | CentraleSupelec Deep G-CRF | 2016-03-28 |
| EdgeNeXt: Efficiently Amalgamated CNN-Transformer Architecture for Mobile Vision Applications | ✓ Link | 80.2% | 8.7G | 6.5M | EdgeNeXt | 2022-06-21 |
| DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs | ✓ Link | 79.7% | | | DeepLab-CRF (ResNet-101) | 2016-06-02 |
| Waterfall Atrous Spatial Pooling Architecture for Efficient Semantic Segmentation | ✓ Link | 79.6% | | | WASPnet-CRF (ours) | 2019-12-06 |
| Light-Weight RefineNet for Real-Time Semantic Segmentation | ✓ Link | 79.2% | | | Light-Weight-RefineNet-MobileNet-v2 | 2018-10-08 |
| The Lovász-Softmax loss: A tractable surrogate for the optimization of the intersection-over-union measure in neural networks | ✓ Link | 79.00% | | | Deeplab-v2 with Lovasz-Softmax loss | 2017-05-24 |
| The Lovász-Softmax loss: A tractable surrogate for the optimization of the intersection-over-union measure in neural networks | ✓ Link | 79.0% | | | Deeplab-v2 + Lovász-Softmax | 2017-05-24 |
| Conditional Random Fields as Recurrent Neural Networks | ✓ Link | 74.7% | | | CRF-RNN | 2015-02-11 |
| Simple Does It: Weakly Supervised Instance and Semantic Segmentation | | 72.8% | | | SID | 2016-03-24 |
| Semantic Image Segmentation with Deep Convolutional Nets and Fully Connected CRFs | ✓ Link | 71.6% | | | DeepLab-MSc-CRF-LargeFOV (VGG-16) | 2014-12-22 |
| ParseNet: Looking Wider to See Better | ✓ Link | 69.8% | | | ParseNet | 2015-06-15 |
| Efficient Yet Deep Convolutional Neural Networks for Semantic Segmentation | ✓ Link | 69% | | | Dilated FCN-2s VGG19 | 2017-07-26 |
| ESPNetv2: A Light-weight, Power Efficient, and General Purpose Convolutional Neural Network | ✓ Link | 68.0% | | | ESPNetv2 | 2018-11-28 |
| Multi-Scale Context Aggregation by Dilated Convolutions | ✓ Link | 67.6% | | | Dilated Convolutions | 2015-11-23 |
| DiCENet: Dimension-wise Convolutions for Efficient Networks | ✓ Link | 67.31% | | | DiCENet | 2019-06-08 |
| Reliability Does Matter: An End-to-End Weakly Supervised Semantic Segmentation Approach | ✓ Link | 66.5 | | | RRM | 2019-11-19 |
| Self-Supervised Difference Detection for Weakly-Supervised Semantic Segmentation | ✓ Link | 65.5 | | | SSDD | 2019-11-04 |
| BoxSup: Exploiting Bounding Boxes to Supervise Convolutional Networks for Semantic Segmentation | | 64.6% | | | BoxSup | 2015-03-05 |
| Find it if You Can: End-to-End Adversarial Erasing for Weakly-Supervised Semantic Segmentation | ✓ Link | 63.8% | | | PSA w/ EADER DeepLab (Xception-65) | 2020-11-09 |
| ESPNet: Efficient Spatial Pyramid of Dilated Convolutions for Semantic Segmentation | ✓ Link | 63.01% | | | ESPNet | 2018-03-19 |
| Fully Convolutional Networks for Semantic Segmentation | ✓ Link | 62.2% | | | FCN (VGG-16) | 2014-11-14 |
| Exploiting saliency for object segmentation from image level labels | | 56.7% | | | G2 | 2017-01-28 |
| Simultaneous Detection and Segmentation | | 51.6% | | | CK | 2014-07-07 |