最新！CVPR 2021 语义分割论文大盘点（39篇论文）

点击下方卡片，关注“CVer”公众号

AI/CV重磅干货，第一时间送达

作者：Amusi  |  来源：CVer

前言

CVer 正式盘点CVPR 2021上各个方向的工作，本篇是热度依然很高的2D语义分割论文大盘点，之前已分享：

• 华人占大半壁江山！CVPR 2021 目标检测论文大盘点（65篇论文）
  

• 最新！CVPR 2021 视觉Transformer论文大盘点（43篇）
  

• 最新！CVPR 2021 OCR领域论文大盘点（22篇）
  

关于更多CVPR 2021的论文和开源代码，可见下面链接：

https://github.com/amusi/CVPR2021-Papers-with-Code

CVPR 2021 2D语义论文（39篇）

Amusi 一共搜集了39篇2D语义分割论文，涉及：通用语义分割、Few-shot/自监督/半监督/域自适应语义分割等方向。

注1：这应该是目前各平台上最新最全面的CVPR 2021 2D语义分割盘点资料，欢迎点赞收藏和分享

注2：3D语义分割、视频目标分割、医学图像分割等检测方向并不在本文范畴，后续将单独分享，敬请期待！

注3：超过一半来自中国的工作，所有论文中，高校以北大、北航、港大、牛津大学、ETH Zurich等为主。

语义分割(Semantic Segmentation)

1. HyperSeg: Patch-wise Hypernetwork for Real-time Semantic Segmentation

• 作者单位: Facebook AI, 巴伊兰大学, 特拉维夫大学

• Homepage: https://nirkin.com/hyperseg/

• Paper: https://openaccess.thecvf.com/content/CVPR2021/papers/Nirkin_HyperSeg_Patch-Wise_Hypernetwork_for_Real-Time_Semantic_Segmentation_CVPR_2021_paper.pdf

• Code: https://github.com/YuvalNirkin/hyperseg

2. Rethinking BiSeNet For Real-time Semantic Segmentation

• 作者单位: 美团

• Paper: https://arxiv.org/abs/2104.13188

• Code: https://github.com/MichaelFan01/STDC-Seg

3. Progressive Semantic Segmentation

• 作者单位: VinAI Research, VinUniversity, 阿肯色大学, 石溪大学

• Paper: https://arxiv.org/abs/2104.03778

• Code: https://github.com/VinAIResearch/MagNet

4. Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers

• 作者单位: 复旦大学, 牛津大学, 萨里大学, 腾讯优图, Facebook AI

• Homepage: https://fudan-zvg.github.io/SETR

• Paper: https://arxiv.org/abs/2012.15840

• Code: https://github.com/fudan-zvg/SETR

5. Capturing Omni-Range Context for Omnidirectional Segmentation

• 作者单位: 卡尔斯鲁厄理工学院, 卡尔·蔡司, 华为

• Paper: https://arxiv.org/abs/2103.05687

• Code: None

6. Learning Statistical Texture for Semantic Segmentation

• 作者单位: 北航, 商汤科技

• Paper: https://arxiv.org/abs/2103.04133

• Code: None

7. InverseForm: A Loss Function for Structured Boundary-Aware Segmentation

• 作者单位: 高通AI研究院

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Borse_InverseForm_A_Loss_Function_for_Structured_Boundary-Aware_Segmentation_CVPR_2021_paper.html

• Code: None

8. DCNAS: Densely Connected Neural Architecture Search for Semantic Image Segmentation

• 作者单位: Joyy Inc, 快手, 北航等

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Zhang_DCNAS_Densely_Connected_Neural_Architecture_Search_for_Semantic_Image_Segmentation_CVPR_2021_paper.html

• Code: None

弱监督语义分割

9. Railroad Is Not a Train: Saliency As Pseudo-Pixel Supervision for Weakly Supervised Semantic Segmentation

• 作者单位: 延世大学, 成均馆大学

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Lee_Railroad_Is_Not_a_Train_Saliency_As_Pseudo-Pixel_Supervision_for_CVPR_2021_paper.html

• Code: https://github.com/halbielee/EPS

10. Background-Aware Pooling and Noise-Aware Loss for Weakly-Supervised Semantic Segmentation

• 作者单位: 延世大学

• Homepage:  https://cvlab.yonsei.ac.kr/projects/BANA/

• Paper: https://arxiv.org/abs/2104.00905

• Code: None

11. Non-Salient Region Object Mining for Weakly Supervised Semantic Segmentation

• 作者单位: 南京理工大学, MBZUAI, 电子科技大学, 阿德莱德大学, 悉尼科技大学

• Paper: https://arxiv.org/abs/2103.14581

• Code: https://github.com/NUST-Machine-Intelligence-Laboratory/nsrom

12. Embedded Discriminative Attention Mechanism for Weakly Supervised Semantic Segmentation

• 作者单位: 北京理工大学, 美团

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Wu_Embedded_Discriminative_Attention_Mechanism_for_Weakly_Supervised_Semantic_Segmentation_CVPR_2021_paper.html

• Code: https://github.com/allenwu97/EDAM

13. BBAM: Bounding Box Attribution Map for Weakly Supervised Semantic and Instance Segmentation

• 作者单位: 首尔大学

• Paper: https://arxiv.org/abs/2103.08907

• Code: https://github.com/jbeomlee93/BBAM

半监督语义分割

14. Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision

• 作者单位: 北京大学, 微软亚洲研究院

• Paper: https://arxiv.org/abs/2106.01226

• Code: https://github.com/charlesCXK/TorchSemiSeg

15. Semi-supervised Domain Adaptation based on Dual-level Domain Mixing for Semantic Segmentation

• 作者单位: 华为, 大连理工大学, 北京大学

• Paper: https://arxiv.org/abs/2103.04705

• Code: None

16. Semi-Supervised Semantic Segmentation With Directional Context-Aware Consistency

• 作者单位: 香港中文大学, 思谋科技, 牛津大学

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Lai_Semi-Supervised_Semantic_Segmentation_With_Directional_Context-Aware_Consistency_CVPR_2021_paper.html

• Code: None

17. Semantic Segmentation With Generative Models: Semi-Supervised Learning and Strong Out-of-Domain Generalization

• 作者单位: NVIDIA, 多伦多大学, 耶鲁大学, MIT, Vector Institute

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Li_Semantic_Segmentation_With_Generative_Models_Semi-Supervised_Learning_and_Strong_Out-of-Domain_CVPR_2021_paper.html

• Code: https://nv-tlabs.github.io/semanticGAN/

18. Three Ways To Improve Semantic Segmentation With Self-Supervised Depth Estimation

• 作者单位: ETH Zurich, 伯恩大学, 鲁汶大学

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Hoyer_Three_Ways_To_Improve_Semantic_Segmentation_With_Self-Supervised_Depth_Estimation_CVPR_2021_paper.html

• Code: https://github.com/lhoyer/improving_segmentation_with_selfsupervised_depth

域自适应语义分割

19. Cluster, Split, Fuse, and Update: Meta-Learning for Open Compound Domain Adaptive Semantic Segmentation

• 作者单位: ETH Zurich, 鲁汶大学, 电子科技大学

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Gong_Cluster_Split_Fuse_and_Update_Meta-Learning_for_Open_Compound_Domain_CVPR_2021_paper.html

• Code: None

20. Source-Free Domain Adaptation for Semantic Segmentation

• 作者单位: 华东师范大学

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Liu_Source-Free_Domain_Adaptation_for_Semantic_Segmentation_CVPR_2021_paper.html

• Code: None

21. Uncertainty Reduction for Model Adaptation in Semantic Segmentation

• 作者单位: Idiap Research Institute, EPFL, 日内瓦大学

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/S_Uncertainty_Reduction_for_Model_Adaptation_in_Semantic_Segmentation_CVPR_2021_paper.html

• Code: https://git.io/JthPp

22. Self-Supervised Augmentation Consistency for Adapting Semantic Segmentation

• 作者单位: 达姆施塔特工业大学, hessian.AI

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Araslanov_Self-Supervised_Augmentation_Consistency_for_Adapting_Semantic_Segmentation_CVPR_2021_paper.html

• Code: https://github.com/visinf/da-sac

23. RobustNet: Improving Domain Generalization in Urban-Scene Segmentation via Instance Selective Whitening

• 作者单位: LG AI研究院, KAIST等

• Paper: https://arxiv.org/abs/2103.15597

• Code: https://github.com/shachoi/RobustNet

24. Coarse-to-Fine Domain Adaptive Semantic Segmentation with Photometric Alignment and Category-Center Regularization

• 作者单位: 香港大学, 深睿医疗

• Paper: https://arxiv.org/abs/2103.13041

• Code: None

25. MetaCorrection: Domain-aware Meta Loss Correction for Unsupervised Domain Adaptation in Semantic Segmentation

• 作者单位: 香港城市大学, 百度

• Paper: https://arxiv.org/abs/2103.05254

• Code: https://github.com/cyang-cityu/MetaCorrection

26. Multi-Source Domain Adaptation with Collaborative Learning for Semantic Segmentation

• 作者单位: 华为云, 华为诺亚, 大连理工大学

• Paper: https://arxiv.org/abs/2103.04717

• Code: None

27. Prototypical Pseudo Label Denoising and Target Structure Learning for Domain Adaptive Semantic Segmentation

• 作者单位: 中国科学技术大学, 微软亚洲研究院

• Paper: https://arxiv.org/abs/2101.10979

• Code: https://github.com/microsoft/ProDA

28. DANNet: A One-Stage Domain Adaptation Network for Unsupervised Nighttime Semantic Segmentation

• 作者单位: 南卡罗来纳大学, 天远视科技

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Wu_DANNet_A_One-Stage_Domain_Adaptation_Network_for_Unsupervised_Nighttime_Semantic_CVPR_2021_paper.html

• Code: https://github.com/W-zx-Y/DANNet

Few-Shot语义分割

29. Scale-Aware Graph Neural Network for Few-Shot Semantic Segmentation

• 作者单位: MBZUAI, IIAI, 哈工大

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Xie_Scale-Aware_Graph_Neural_Network_for_Few-Shot_Semantic_Segmentation_CVPR_2021_paper.html

• Code: None

30. Anti-Aliasing Semantic Reconstruction for Few-Shot Semantic Segmentation

• 作者单位: 国科大, 清华大学

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Liu_Anti-Aliasing_Semantic_Reconstruction_for_Few-Shot_Semantic_Segmentation_CVPR_2021_paper.html

• Code: https://github.com/Bibkiller/ASR

无监督语义分割

31. PiCIE: Unsupervised Semantic Segmentation Using Invariance and Equivariance in Clustering

• 作者单位: UT-Austin, 康奈尔大学

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Cho_PiCIE_Unsupervised_Semantic_Segmentation_Using_Invariance_and_Equivariance_in_Clustering_CVPR_2021_paper.html

• Code: https:// github.com/janghyuncho/PiCIE

视频语义分割

32. VSPW: A Large-scale Dataset for Video Scene Parsing in the Wild

• 作者单位: 浙江大学, 百度, 悉尼科技大学

• Homepage: https://www.vspwdataset.com/

• Paper: https://www.vspwdataset.com/CVPR2021__miao.pdf

• GitHub: https://github.com/sssdddwww2/vspw_dataset_download

其它

33. Continual Semantic Segmentation via Repulsion-Attraction of Sparse and Disentangled Latent Representations

• 作者单位: 帕多瓦大学

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Michieli_Continual_Semantic_Segmentation_via_Repulsion-Attraction_of_Sparse_and_Disentangled_Latent_CVPR_2021_paper.html

• Code: https://lttm.dei.unipd.it/paper_data/SDR/

34. Exploit Visual Dependency Relations for Semantic Segmentation

• 作者单位: 伊利诺伊大学芝加哥分校

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Liu_Exploit_Visual_Dependency_Relations_for_Semantic_Segmentation_CVPR_2021_paper.html

• Code: None

35. Revisiting Superpixels for Active Learning in Semantic Segmentation With Realistic Annotation Costs

• 作者单位: Institute for Infocomm Research, 新加坡国立大学

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Cai_Revisiting_Superpixels_for_Active_Learning_in_Semantic_Segmentation_With_Realistic_CVPR_2021_paper.html

• Code: None

36. PLOP: Learning without Forgetting for Continual Semantic Segmentation

• 作者单位: 索邦大学, Heuritech, Datakalab, Valeo.ai

• Paper: https://arxiv.org/abs/2011.11390

• Code: https://github.com/arthurdouillard/CVPR2021_PLOP

37. 3D-to-2D Distillation for Indoor Scene Parsing

• 作者单位: 香港中文大学, 香港大学

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Liu_3D-to-2D_Distillation_for_Indoor_Scene_Parsing_CVPR_2021_paper.html

• Code: None

38. Bidirectional Projection Network for Cross Dimension Scene Understanding

• 作者单位: 香港中文大学, 牛津大学等

• Paper(Oral): https://arxiv.org/abs/2103.14326

• Code: https://github.com/wbhu/BPNet

39. PointFlow: Flowing Semantics Through Points for Aerial Image Segmentation

• 作者单位: 北京大学, 中科院, 国科大, ETH Zurich, 商汤科技等

• Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Li_PointFlow_Flowing_Semantics_Through_Points_for_Aerial_Image_Segmentation_CVPR_2021_paper.html

• Code: https://github.com/lxtGH/PFSegNets

上述39篇语义分割论文下载
后台回复：CVPR2021，即可下载上述论文PDF
CVPR和Transformer资料下载

后台回复：CVPR2021，即可下载CVPR 2021论文和代码开源的论文合集
后台回复：Transformer综述，即可下载最新的两篇Transformer综述PDF
CVer-语义分割交流群成立
扫码添加CVer助手，可申请加入CVer-语义分割 微信交流群，方向已涵盖：目标检测、图像分割、目标跟踪、人脸检测&识别、OCR、姿态估计、超分辨率、SLAM、医疗影像、Re-ID、GAN、NAS、深度估计、自动驾驶、强化学习、车道线检测、模型剪枝&压缩、去噪、去雾、去雨、风格迁移、遥感图像、行为识别、视频理解、图像融合、图像检索、论文投稿&交流、PyTorch和TensorFlow等群。
一定要备注：研究方向+地点+学校/公司+昵称（如语义分割+上海+上交+卡卡），根据格式备注，可更快被通过且邀请进群
▲长按加小助手微信，进交流群▲点击上方卡片，关注CVer公众号
整理不易，请给CVer点赞和在看