├── Regularization.pdf ├── Supervised-segmentation.pdf ├── Eye-diasease.md ├── continue-learning.md ├── NAS.md ├── weakly-labelled.md ├── ICLR2020.md ├── ReadMe.md └── Unsupervised-learning.md /Regularization.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/emma-sjwang/Reading_list/HEAD/Regularization.pdf -------------------------------------------------------------------------------- /Supervised-segmentation.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/emma-sjwang/Reading_list/HEAD/Supervised-segmentation.pdf -------------------------------------------------------------------------------- /Eye-diasease.md: -------------------------------------------------------------------------------- 1 | ## Glaucoma 2 | - **[]** [Real-World Multi-Domain Data Applications for Generalizations to Clinical Settings](https://arxiv.org/pdf/2007.12672.pdf) ： self-supervised learning, contrastive loss 3 | -------------------------------------------------------------------------------- /continue-learning.md: -------------------------------------------------------------------------------- 1 | # continue Learning 2 | 3 | ## Medical 4 | - **[MICCAI/2020]** [Continual Learning of New Diseases with Dual Distillation and Ensemble Strategy](https://link.springer.com/chapter/10.1007/978-3-030-59710-8_17) : multi-domain, channel re-weighting 5 | -------------------------------------------------------------------------------- /NAS.md: -------------------------------------------------------------------------------- 1 | # Network Architecture Search (NAS) 2 | 3 | 4 | > - Method [Type/Venue] Name+link: introduction 5 | 6 | ## Medical Applications 7 | - **[MICCAI20]** [UXNet: Searching Multi-level Feature Aggregation for 3D Medical Image Segmentation](https://arxiv.org/abs/2009.07501): block + scale level searching 8 | 9 | ### 2020 10 | - **SGAS [Gradient/CVPR]** [SGAS: Sequential Greedy Architecture Search](https://www.deepgcns.org/auto/sgas): 基于[DARTS](#DARTS)框架,使用贪心的方式来搜索局部最优的architecture, 可以用于CNN和GCN. 事先定义好多少层feature,然后搜索的是feature之间的*卷积类型*或者是*skip-connection*, 并不涉及到architecture的大小扩增. 11 | 12 | ### 2019 13 | - **DARTS [Gradient/ICLR]** [DARTS: Differentiable Architecture Search](https://openreview.net/pdf?id=S1eYHoC5FX): -------------------------------------------------------------------------------- /weakly-labelled.md: -------------------------------------------------------------------------------- 1 | # Weakly-labelled 2 | 3 | ## Medical 4 | - **[ICCV/2019]** [Prior-aware Neural Network for Partially-SupervisedMulti-Organ Segmentation](https://openaccess.thecvf.com/content_ICCV_2019/papers/Zhou_Prior-Aware_Neural_Network_for_Partially-Supervised_Multi-Organ_Segmentation_ICCV_2019_paper.pdf): distribution alignment. Organ segmentation 5 | - **[MIDL/2019]** [Learning with Multitask Adversaries using Weakly Labelled Data for Semantic Segmentation in Retinal Images](https://openreview.net/pdf?id=HJe6f0BexN) : structure segmentation and lesion segmentation 6 | - **[MIA/2020]** [Learning joint segmentation of tissues and brain lesions from task-specific hetero-modal domain-shifted datasets](https://www.sciencedirect.com/science/article/pii/S1361841520302267) : segmentation 7 | 8 | - [TMI/2020] [Deep Mining External Imperfect Data for ChestX-ray Disease Screening](https://arxiv.org/pdf/2006.03796.pdf): 有一部分label是两个数据集之间共享的,然后每个数据集都有自己特殊的一部分label. : classification 9 | - **[MICCAI/2020]** [Multi-organ Segmentation via Co-training Weight-Averaged Models from Few-Organ Datasets](https://link.springer.com/chapter/10.1007/978-3-030-59719-1_15): multi-organ segmentation similar to Yixiao Ge. Two semi-framework 10 | - **[arxiv/2020]** [Universal Lesion Detection by Learning fromMultiple Heterogeneously Labeled Datasets](https://arxiv.org/pdf/2005.13753.pdf): lesion segmentation on dataset [DeepLesion], similar to negative mining 11 | - **[arxiv/2020]** [Learning from Multiple Datasets withHeterogeneous and Partial Labelsfor Universal Lesion Detection in CT](https://arxiv.org/pdf/2009.02577.pdf): false positive reduction. multi kind of lesions detection 12 | - **[ICCV/2019]** [Stochastic Filter Groups for Multi-Task CNNs:Learning Specialist and Generalist Convolution Kernels](https://openaccess.thecvf.com/content_ICCV_2019/papers/Bragman_Stochastic_Filter_Groups_for_Multi-Task_CNNs_Learning_Specialist_and_Generalist_ICCV_2019_paper.pdf): organ segmentation, network architecture group design 13 | - **[MICCAI workshop/2020]** [Continual Class Incremental Learningfor CT Thoracic Segmentation](https://arxiv.org/abs/2008.05557): multi-organ segmentation. continual learning. 14 | - **[MIML/2020]** [Importance Driven Continual Learning forSegmentation Across Domains](https://link.springer.com/chapter/10.1007/978-3-030-59861-7_43): brain segmentation 15 | - **[AAAi/2020]** [Learning from Weak-Label Data: A Deep Forest Expedition](http://www.lamda.nju.edu.cn/yangl/paper/AAAI.pdf): multi-layer random forest with label correction (weak label estimation) 16 | 17 | 18 | 19 | # Fundus multi-task 20 | ## DR+DME 21 | - **[ISBI/2020]** [SUNET: A LESION REGULARIZED MODEL FOR SIMULTANEOUS DIABETICRETINOPATHY AND DIABETIC MACULAR EDEMA GRADING](https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9098673): network block design 22 | - **[arxiv/2020]** [Predicting Risk of Developing Diabetic Retinopathy using Deep Learning](https://arxiv.org/abs/2008.04370) :Google, large scale dataset using inceptation-v3 -------------------------------------------------------------------------------- /ICLR2020.md: -------------------------------------------------------------------------------- 1 | # [ICLR 2020](https://openreview.net/group?id=ICLR.cc/2020/Conference) Interesting papers 2 | >There is nerver nothing we can do. 3 | 4 | 5 | ## Generative Model 6 | 7 | * [**promising results!**] [ U-GAT-IT: Unsupervised Generative Attentional Networks with Adaptive Layer-Instance Normalization for Image-to-Image Translation ](https://openreview.net/pdf?id=BJlZ5ySKPH) [[Tensorflow](https://github.com/taki0112/UGATIT)]: Style transfer 8 | 9 | * [**Discriminator**] [Real or not real, that is the question](https://openreview.net/pdf?id=B1lPaCNtPB) [[PyTorch](https://github.com/kam1107/RealnessGAN)]: Change the output of discriminator to a distribution with a vector size of 8 (followed by KL loss), which is similar to the extended soft label. 10 | 11 | * [**Understanding generative model**] [Do deep generative models know what they don't know](https://openreview.net/forum?id=H1xwNhCcYm) 12 | 13 | * :star::star::star::star::star: [**Robust GAN**] [Optimal strategies against generative attacks](https://openreview.net/pdf?id=BkgzMCVtPB) [[PyTorch](https://github.com/roymor1/OptimalStrategiesAgainstGenerativeAttacks)]: GAN in the middle networks. Find the optimal solution to against attacks, with analysis of how to attack with leaked samples. 14 | 15 | * [**Loss function**] [Improving Adversarial Robustness Requires Revisiting Misclassified Examples](https://openreview.net/pdf?id=rklOg6EFwS) : [Mark](https://iclr.cc/virtual/poster_rklOg6EFwS.html) 16 | 17 | * [**Comparison discriminator**] [Self-Adversarial Learning with Comparative Discrimination for Text Generation ](https://openreview.net/pdf?id=B1l8L6EtDS) 18 | 19 | * [Controlling generative models with continuous factors of variations](https://openreview.net/pdf?id=H1laeJrKDB) 20 | 21 | * [**Learning para as loss weight**] [You Only Train Once: Loss-Conditional Training of Deep Networks](https://openreview.net/pdf?id=HyxY6JHKwr) 22 | 23 | * [**Augmentation**] [Adversarial AutoAugment](https://openreview.net/pdf?id=ByxdUySKvS) : RL 24 | 25 | 26 | 27 | 28 | 29 | ## Generalization 30 | 31 | * [Identity Crisis: Memorization and Generalization Under Extreme Overparameterization](https://openreview.net/forum?id=B1l6y0VFPr): FC layers. Low level features show identity ability, while high-level features output constant information. 32 | 33 | 34 | 35 | ## Representation Learning 36 | * [Target-Embedding Autoencoders for Supervised Representation Learning](https://openreview.net/pdf?id=BygXFkSYDH) : VAE usually reconstracts X, but in this paper, it reconstracts Y(Y is high-dimensional). Application: multivariate sequence forecasting 37 | 38 | * :star::star::star::star::star: [**Unsupervised clustering**] [Self-labelling via simultaneous clustering and representation learning ](https://openreview.net/pdf?id=Hyx-jyBFPr) [[PyTorch](https://github.com/yukimasano/self-label)]: random generated labels & optimization & update labels (representation learning) + linear program solving 39 | 40 | * [Interpretation] [Rotation-invariant clustering of neuronal responses in primary visual cortex ](https://openreview.net/pdf?id=rklr9kHFDB) 41 | 42 | * [Multi-view information] [Learning Robust Representations via Multi-View Information Bottleneck ](https://openreview.net/pdf?id=B1xwcyHFDr) 43 | 44 | 45 | 46 | ## Domain Adaptation 47 | 48 | * [Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification](https://openreview.net/forum?id=rJlnOhVYPS) : re-identification, mean teacher, 4 models, triplet loss 49 | 50 | * [Domain Adaptive Multibranch Networks](https://openreview.net/pdf?id=rJxycxHKDS) 51 | 52 | 53 | 54 | 55 | 56 | ## Multi-modal (image/text) 57 | 58 | * [ Latent Normalizing Flows for Many-to-Many Cross-Domain Mappings ](https://openreview.net/pdf?id=SJxE8erKDH) [[PyTorch](https://github.com/visinf/lnfmm)]: image caption 59 | * [From Inference to Generation: End-to-end Fully Self-supervised Generation of Human Face from Speech ](https://openreview.net/pdf?id=H1guaREYPr): using voice to generate face, interesting. 60 | 61 | 62 | 63 | 64 | ## Graph Neural Network 65 | * [Deep Graph Matching Consensus](https://openreview.net/pdf?id=HyeJf1HKvS) [[PyTorch](https://github.com/rusty1s/deep-graph-matching-consensus)]: Idea (neighboors should contain same information in two similar images) is good, which might be used in somewhere else. 66 | * [DropEdge: Towards Deep Graph Convolutional Networks on Node Classification ](https://openreview.net/pdf?id=Hkx1qkrKPr) [[PyTorch](https://github.com/DropEdge/DropEdge)] 67 | 68 | 69 | 70 | ## Self-supervised Learning 71 | 72 | * :star::star::star::star: [ A crical analysis of self-supervision, or what we can learn from a single image](https://openreview.net/pdf?id=B1esx6EYvr) : Self supervision with one image could learn low-level features with high quality 73 | 74 | 75 | 76 | ## Semi-supervised Learning 77 | 78 | * [**Application**] [Automatically Discovering and Learning New Visual Categories with Ranking Statistics](https://openreview.net/pdf?id=BJl2_nVFPB) [[PyTorch](http://www.robots.ox.ac.uk/~vgg/research/auto_novel)]: 应用流也能中ICLR了.. self-supervision + supervised-learning + pesudo label + incremental learning 79 | 80 | * [ Deep Semi-Supervised Anomaly Detection ](https://openreview.net/pdf?id=HkgH0TEYwH) 81 | 82 | * [**Remove noise labelled data to semi-supervised learning**] [ SELF: Learning to Filter Noisy Labels with Self-Ensembling ](https://openreview.net/pdf?id=HkgsPhNYPS) 83 | 84 | * [**Mixmatch upgrade**] [ReMixMatch: Semi-Supervised Learning with Distribution Matching and Augmentation Anchoring](https://openreview.net/pdf?id=HklkeR4KPB) 85 | 86 | * [**Semi-supervised**] [DivideMix: Learning with Noisy Labels as Semi-supervised Learning](https://openreview.net/pdf?id=HJgExaVtwr): filter noise label, then trest them as unlabelled data, apply pseudo label for mixmatch. 87 | 88 | 89 | 90 | ## Weakly-supervised Learning 91 | 92 | * [Weakly Supervised Clustering by Exploiting Unique Class Count](https://openreview.net/pdf?id=B1xIj3VYvr): predict the class No. within one images. 93 | 94 | 95 | 96 | ## Active Learning 97 | 98 | * [certainty and diversity] [Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds](https://openreview.net/pdf?id=ryghZJBKPS) [Oral] 99 | 100 | 101 | 102 | ## Disentangle 103 | 104 | * [Weakly Supervised Disentanglement with Guarantees](https://openreview.net/pdf?id=HJgSwyBKvr) 105 | 106 | 107 | 108 | ## Noise Lable 109 | 110 | * [**Remove noise labelled data to semi-supervised learning**] [ SELF: Learning to Filter Noisy Labels with Self-Ensembling ](https://openreview.net/pdf?id=HkgsPhNYPS) 111 | 112 | * [**Regularization**] [Simple and Effective Regularization Methods for Training on Noisily Labeled Data with Generalization Guarantee](https://openreview.net/pdf?id=Hke3gyHYwH) [[Code](https://drive.google.com/drive/folders/1TDlUuL0I-EzIybjz2pMAgyaYP5F6dq6o)] 113 | 114 | * [**Semi-supervised**] [DivideMix: Learning with Noisy Labels as Semi-supervised Learning](https://openreview.net/pdf?id=HJgExaVtwr): filter noise label, then trest them as unlabelled data, apply pseudo label for mixmatch. 115 | 116 | 117 | 118 | ## Network Architecture 119 | * [**Semantic Segmentation**] [FasterSeg:Searching for Faster Real-time Semantic Segmentation](https://openreview.net/pdf?id=BJgqQ6NYvB) [[PyTorch](https://github.com/TAMU-VITA/FasterSeg)]: network search + teacher/student knowledge distilation 120 | 121 | 122 | 123 | 124 | ## Network Component Upgrade 125 | * [**Activation**] [Enhancing adversarial defence by k-winners-take-all](https://openreview.net/pdf?id=Skgvy64tvr) [[PyTorch](https://github.com/a554b554/kWTA-Activation)]: For improving the robustness, similar to ReLU but with a ratio to keep the values instead of using threshold 0. 126 | 127 | * [**Curriculum Loss**] [ Curriculum Loss: Robust Learning and Generalization against Label Corruption ](https://openreview.net/pdf?id=rkgt0REKwS) 128 | 129 | * [**Normalization**] [Mixup Inference: Better Exploiting Mixup to Defend Adversarial Attacks](https://openreview.net/pdf?id=ByxtC2VtPB) [[PyTorch](https://github.com/P2333/Mixup-Inference)]: Adding permutations to the samples to improve the robustness. 130 | 131 | * [**Optimization**] [Don't Use Large Mini-batches, Use Local SGD](https://openreview.net/pdf?id=B1eyO1BFPr) multi-GPU, before communication, more inference... 132 | 133 | * [**Optimization**] [Large Batch Optimization for Deep Learning: Training BERT in 76 minutes ](https://openreview.net/pdf?id=Syx4wnEtvH) [[TensorFlow](https://github.com/tensorflow/addons/blob/master/tensorflow_addons/optimizers/lamb.py)] 134 | 135 | 136 | 137 | ## Meta-Learning 138 | * [Learning to Balance: Bayesian Meta-Learning for Imbalanced and Out-of-distribution Tasks](https://openreview.net/pdf?id=rkeZIJBYvr) [[Code](https://github.com/haebeom-lee/l2b)][Oral] 139 | 140 | * [ A Meta-Transfer Objective for Learning to Disentangle Causal Mechanisms ](https://openreview.net/pdf?id=ryxWIgBFPS) 141 | 142 | 143 | 144 | 145 | ## Point Cloud 146 | * [Unpaired Point Cloud Completion on Real Scans using Adversarial Training](https://openreview.net/pdf?id=HkgrZ0EYwB) [[TensorFlow](https://github.com/xuelin-chen/pcl2pcl-gan-pub)]: GAN + super-resolution 147 | 148 | 149 | 150 | ## Mobile Network 151 | 152 | * [**Model Compression**] [Once for All: Train One Network and Specialize it for Efficient Deployment](https://openreview.net/pdf?id=HylxE1HKwS) [[PyTorch](https://github.com/mit-han-lab/once-for-all)] 153 | 154 | * -------------------------------------------------------------------------------- /ReadMe.md: -------------------------------------------------------------------------------- 1 | # Reading_list 2 | 3 | *** 4 | 5 | ## Unsupervised learning 6 | ### Domain Generalization 7 | 8 | ### Domain Adaptation 9 | 10 | ### AAE 11 | * **[AAE](https://arxiv.org/pdf/1511.05644.pdf)** Makhzani, Alireza, et al. "Adversarial autoencoders." arXiv preprint arXiv:1511.05644 (2015). 12 | 13 | ### GAN 14 | 15 | ## pooling 16 | * **[ContextLocNet: Context-Aware Deep Network Models for Weakly Supervised Localization](https://link.springer.com/content/pdf/10.1007\%2F978-3-319-46454-1_22.pdf)** 17 | 18 | 19 | ## Weakly Supervised Learning 20 | * **[Weakly Supervised Object Localization with Progressive Domain Adaptation](https://www.cv-foundation.org/openaccess/content_cvpr_2016/app/S15-30.pdf)** classification and detection 2 steps. only image labels. [CVPR 2016] 21 | * **[ICCV 2015 Deep Learning Face Attributes in the Wild ]** 22 | * **[CVPR 2017 Weakly Supervised Cascaded Convolutional Networks](http://openaccess.thecvf.com/content_cvpr_2017/papers/Diba_Weakly_Supervised_Cascaded_CVPR_2017_paper.pdf)** 23 | * **[CVPR 2016 Large Scale Semi-supervised Object Detection using Visual and Semantic Knowledge Transfer](https://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Tang_Large_Scale_Semi-Supervised_CVPR_2016_paper.pdf)** 24 | * **[CVPR 2018 Class Peak Response](Class Peak Response)** 25 | 26 | 27 | ## Semi-supervised 28 | * http://ruder.io/semi-supervised/ 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | ## optic disk and optic cup segmentation 39 | * [Auto-Encoding Variational Bayes](https://arxiv.org/pdf/1312.6114.pdf) 40 | ### segmentation 41 | * a survey 2017 almost traditional methods [Segmentation Techniques for Computer-Aided Diagnosis of Glaucoma: A Review](https://link.springer.com/content/pdf/10.1007\%2F978-981-10-8569-7_18.pdf) :star::star: 42 | 43 | * based on structural and gray level properties [Segmentation of optic disk and optic cup from digital fundus images for the assessment of glaucoma](https://ac.els-cdn.com/S1746809415001512/1-s2.0-S1746809415001512-main.pdf?_tid=28900699-9a58-4929-8c0c-6268f84bdaf4&acdnat=1529743095_1f81babc1977bd510cafed242e7f2ac3) no deep learning but very useful ideas. :star::star::star::star: 44 | 45 | * convolutional filter + entropy sampling + convex hull transformation [Glaucoma detection using entropy sampling and ensemble learning for automatic optic cup and disc segmentation 46 | ](https://ac.els-cdn.com/S0895611116300775/1-s2.0-S0895611116300775-main.pdf?_tid=6678eb0e-7fcd-4a4e-89f6-fe8cad5730f4&acdnat=1529756396_01824474b64b63bab5be415a9f933f00) too complicated :star::star::star: 47 | 48 | * semi-supervised [Semi-supervised Segmentation of Optic Cup in Retinal Fundus Images Using Variational Autoencoder](https://link.springer.com/content/pdf/10.1007\%2F978-3-319-66185-8_9.pdf) MICCAI 2017 :star::star::star::star: 49 | 50 | * [Joint Optic Disc and Cup Segmentation Based on Multi-label Deep Network and Polar Transformation](https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8252743) 2018 TMI :star::star::star::star::star: 51 | * [Deep Retinal Image Understanding](https://arxiv.org/pdf/1609.01103.pdf) 2016 MICCAI :star::star::star::star::star: 52 | 53 | 54 | 55 | ### domain adaptation 56 | * [Adversarial Discriminative Domain Adaptation](http://openaccess.thecvf.com/content_cvpr_2017/papers/Tzeng_Adversarial_Discriminative_Domain_CVPR_2017_paper.pdf) CVPR 2017 :star::star::star::star: 57 | * [Adversarial Feature Augmentation for Unsupervised Domain Adaptation](https://arxiv.org/abs/1711.08561) CVPR 2018. :star::star::star::star::star: 58 | * [Unsupervised Pixel?Level Domain Adaptation with Generative Adversarial Networks](http://openaccess.thecvf.com/content_cvpr_2017/papers/Bousmalis_Unsupervised_Pixel-Level_Domain_CVPR_2017_paper.pdf) CVPR 2017 :star::star::star::star: 59 | * [Boosting Domain Adaptation by Discovering Latent Domains](https://arxiv.org/abs/1805.01386) CVPR 2018.oral 60 | * [Learning to Adapt Structured Output Space for Semantic Segmentation](). CVPR 2018. GTA IoU: 42.4\% SYNTHIA IoU: 46.7\% 61 | * [Conditional Generative Adversarial Network for Structured Domain Adaptation]() CVPR 2018. GTA IoU 44.5\%. SYNTHIA IoU: 41.2\% 62 | * [Unsupervised Domain Adaptation with Similarity Learning]() for classification 63 | 64 | [Domain Adaptation for Segmentation] 65 | 66 | *[CycleGAN Multimodal, shape constraints ](https://arxiv.org/pdf/1802.09655.pdf) CVPR2018 :star::star::star::star::star: 67 | 68 | 69 | ### patch-level reinforesment learning 70 | 71 | 72 | *** 73 | ## MIML (multi-instance multi label learning) 74 | * **Deep MIML** Feng, Ji, and Zhi-Hua Zhou. "Deep MIML Network." AAAI. 2017. 75 | 76 | ### MIL 77 | * **[survey](https://www.sciencedirect.com/science/article/pii/S0031320317304065)** Carbonneau, Marc-André, et al. "Multiple instance learning: A survey of problem characteristics and applications." Pattern Recognition (2017).  [another link here](https://arxiv.org/pdf/1612.03365.pdf) :star::star::star::star::star: 78 | 79 | #### Instance Space Method 80 | * **[EM_DD](http://papers.nips.cc/paper/1959-em-dd-an-improved-multiple-instance-learning-technique.pdf)** Zhang, Qi, and Sally A. Goldman. "EM-DD: An improved multiple-instance learning technique." Advances in neural information processing systems. 2002. [Implement](http://lamda.nju.edu.cn/code_MIL-Ensemble.ashx) 81 | :star::star::star::star: 82 | * **[MI_SVM](http://papers.nips.cc/paper/2232-support-vector-machines-for-multiple-instance-learning.pdf)** Andrews, Stuart, Ioannis Tsochantaridis, and Thomas Hofmann. "Support vector machines for multiple-instance learning." Advances in neural information processing systems. 2003. :star::star::star::star: 83 | * **[MILBoost](http://papers.nips.cc/paper/2926-multiple-instance-boosting-for-object-detection.pdf)** Zhang, Cha, John C. Platt, and Paul A. Viola. "Multiple instance boosting for object detection." Advances in neural information processing systems. 2006. :star::star::star::star: 84 | 85 | #### Bag Space Method 86 | * **[Diverse Density(DD)](http://lis.csail.mit.edu/pubs/tlp/maron98framework.pdf)** Maron, Oded, and Tomás Lozano-Pérez. "A framework for multiple-instance learning." Advances in neural information processing systems. 1998. [Implement](http://lamda.nju.edu.cn/code_MIL-Ensemble.ashx) :star::star::star::star: 87 | * **[citation kNN](http://cogprints.org/2124/3/wang_ICML2000.pdf)** Wang, Jun, and Jean-Daniel Zucker. "Solving multiple-instance problem: A lazy learning approach." (2000): 1119-1125. [Implement](http://lamda.nju.edu.cn/code_MIL-Ensemble.ashx) 88 | * **[MInd](https://www.sciencedirect.com/science/article/pii/S0031320314002817)** Cheplygina, Veronika, David MJ Tax, and Marco Loog. "Multiple instance learning with bag dissimilarities." Pattern Recognition 48.1 (2015): 264-275. 89 | * **[CCE](https://link.springer.com/content/pdf/10.1007%2Fs10115-006-0029-3.pdf)** Zhou, Zhi-Hua, and Min-Ling Zhang. "Solving multi-instance problems with classifier ensemble based on constructive clustering." Knowledge and Information Systems 11.2 (2007): 155-170. [implement](http://lamda.nju.edu.cn/code_CCE.ashx) 90 | 91 | * **[MILES](http://ieeexplore.ieee.org/abstract/document/1717454/)** Chen, Yixin, Jinbo Bi, and James Ze Wang. "MILES: Multiple-instance learning via embedded instance selection." IEEE Transactions on Pattern Analysis and Machine Intelligence 28.12 (2006): 1931-1947. 92 | * **[NSK-SVM](http://sci2s.ugr.es/keel/pdf/algorithm/congreso/2002-Gartner-ICML.pdf)** Gärtner, Thomas, et al. "Multi-instance kernels." ICML. Vol. 2. 2002. 93 | * **[mi-Graph](http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.149.6808&rep=rep1&type=pdf)** Zhou, Zhi-Hua, Yu-Yin Sun, and Yu-Feng Li. "Multi-instance learning by treating instances as non-iid samples." Proceedings of the 26th annual international conference on machine learning. ACM, 2009. [implement](http://lamda.nju.edu.cn/code_miGraph.ashx) 94 | * **[BoW-SVM]()** 95 | * **[EMD-SVM](https://link.springer.com/content/pdf/10.1023%2FA%3A1026543900054.pdf)** Rubner, Yossi, Carlo Tomasi, and Leonidas J. Guibas. "The earth mover's distance as a metric for image retrieval." International journal of computer vision 40.2 (2000): 99-121. 96 | 97 | #### Ranking 98 | * **** Fast bundle algorithm for multiple-instance learning 99 | * **** Multiple-instance ranking: Learning to rank images for image retrieval 100 | 101 | #### others 102 | * **[MIL pooling layer](https://academic.oup.com/bioinformatics/article/32/12/i52/2288769)** Kraus, Oren Z., Jimmy Lei Ba, and Brendan J. Frey. "Classifying and segmenting microscopy images with deep multiple instance learning." Bioinformatics 32.12 (2016): i52-i59.   :star::star::star::star: 103 | 104 | * **[ multi-instane neural network](https://lirias.kuleuven.be/bitstream/123456789/133224/1/31670.pdf)** Ramon, Jan, and Luc De Raedt. "Multi instance neural networks." Proceedings of the ICML-2000 workshop on attribute-value and relational learning. 2000. :star::star::star: 105 | 106 | 107 | 108 | * **[ML-KNN](https://ac.els-cdn.com/S0031320307000027/1-s2.0-S0031320307000027-main.pdf?_tid=4aca996e-88bb-4a31-8efc-b590364adbd2&acdnat=1521359388_bb21b8697481230d67ebf257245dad8a)** Zhang, Min-Ling, and Zhi-Hua Zhou. "ML-KNN: A lazy learning approach to multi-label learning." Pattern recognition 40.7 (2007): 2038-2048. 109 | 110 | #### MIL in Deep Learning 111 | * **[Multi-Instance Deep Learning: Discover Discriminative Local Anatomies for Bodypart Recognition](http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7398101)** Yan, Zhennan, et al. "Multi-instance deep learning: Discover discriminative local anatomies for bodypart recognition." IEEE transactions on medical imaging 35.5 (2016): 1332-1343. 112 | * **[MILCNN](https://arxiv.org/pdf/1610.03155.pdf)** Sun, Miao, et al. "Multiple instance learning convolutional neural networks for object recognition." Pattern Recognition (ICPR), 2016 23rd International Conference on. IEEE, 2016. 113 | * **[Attention Deep MIL](https://arxiv.org/pdf/1802.04712.pdf)** Ilse, Maximilian, Jakub M. Tomczak, and Max Welling. "Attention-based Deep Multiple Instance Learning." arXiv preprint arXiv:1802.04712 (2018). :star::star::star::star::star::star: 114 | * **[MINN](https://ac.els-cdn.com/S0031320317303382/1-s2.0-S0031320317303382-main.pdf?_tid=a6cd7eba-7151-4cf6-9bae-7920c3e0ac75&acdnat=1521535813_58887e0c10d507eeab20ecc7e9b012e5)** Wang, Xinggang, et al. "Revisiting multiple instance neural networks." Pattern Recognition 74 (2018): 15-24. :star::star::star::star::star: 115 | 116 | ### SEMI-SUPERVISED LEARNING 117 | * **[unsupervised loss function](https://arxiv.org/pdf/1606.04586.pdf)** Sajjadi, Mehdi, Mehran Javanmardi, and Tolga Tasdizen. "Regularization with stochastic transformations and perturbations for deep semi-supervised learning." Advances in Neural Information Processing Systems. 2016. 118 | * **[self-ensembling](https://arxiv.org/pdf/1610.02242.pdf)** Laine, Samuli, and Timo Aila. "Temporal ensembling for semi-supervised learning." arXiv preprint arXiv:1610.02242 (2016). 119 | 120 | 121 | ### Loss Function 122 | * **[loss function based on probability map](https://arxiv.org/abs/1804.01793)** Jetley, Saumya, Naila Murray, and Eleonora Vig. "End-to-end saliency mapping via probability distribution prediction." Proceedings of Computer Vision and Pattern Recognition 2016 (2016): 5753-5761. 123 | * **[L-GM loss for image classification](https://arxiv.org/abs/1803.02988)** Wan, Weitao, et al. "Rethinking Feature Distribution for Loss Functions in Image Classification." arXiv preprint arXiv:1803.02988 (2018). [CVPR 2018]() :star::star::star::star::star: [implement](https://github.com/WeitaoVan/L-GM-loss) 124 | * **[Crystal Loss(softmax+l\_2 norm)](https://arxiv.org/pdf/1804.01159.pdf)** Crystal Loss and Quality Pooling for Unconstrained Face Verification and Recognition. submitted to TPAMI 2018. [previous version](https://arxiv.org/abs/1703.09507) 125 | * **[ring loss for face recognation](https://arxiv.org/abs/1803.00130)** Zheng, Yutong, Dipan K. Pal, and Marios Savvides. "Ring loss: Convex Feature Normalization for Face Recognition." arXiv preprint arXiv:1803.00130 (2018). [CVPR 2018]() :star::star::star::star::star: [implement](https://github.com/Paralysis/ringloss) 126 | * **[center loss](https://ydwen.github.io/papers/WenECCV16.pdf)** Wen, Yandong, et al. "A discriminative feature learning approach for deep face recognition." European Conference on Computer Vision. Springer, Cham, 2016. :star::star::star::star::star: 127 | 128 | ### Feature Extraction 129 | * **[Adaptive forward-backward greedy algorithm](http://ieeexplore.ieee.org/abstract/document/5895111/)** Zhang, Tong. "Adaptive forward-backward greedy algorithm for learning sparse representations." IEEE transactions on information theory 57.7 (2011): 4689-4708. 130 | 131 | 132 | 133 | -------------------------------------------------------------------------------- /Unsupervised-learning.md: -------------------------------------------------------------------------------- 1 | # Unsupervised Learning 2 | 3 | ## Domain Generalization (DG) 4 | - **[CVPR/2020]** [Multi-Domain Learning for Accurate and Few-Shot Color Constancy](https://openaccess.thecvf.com/content_CVPR_2020/papers/Xiao_Multi-Domain_Learning_for_Accurate_and_Few-Shot_Color_Constancy_CVPR_2020_paper.pdf) : multi-domain, channel re-weighting 5 | - **[arxiv]** [Domain Generalization via Semi-supervised Meta Learning](https://arxiv.org/pdf/2009.12658.pdf): supervised loss, semi-loss, alignment loss. 6 | - **LDDG [NIPS/2020]** [Domain Generalization for Medical Imaging Classification with Linear-Dependency Regularization](https://arxiv.org/abs/2009.12829) 7 | - **EISNet [ECCV/2020]** [Learning from Extrinsic and Intrinsic Supervisions for Domain Generalization](https://github.com/EmmaW8/EISNet) 8 | - **DMG [ECCV/2020]** [Learning to Balance Specificity and Invariance for In and Out of Domain Generalization](https://arxiv.org/pdf/2008.12839.pdf): domain specific masks. similar to attention map **DATA**: PACS, DomainNet 9 | - **[ECCV/2020(oral)]** [Self-Challenging Improves Cross-Domain Generalization](https://arxiv.org/abs/2007.02454) Interesting. mute activate node. 10 | - **[ECCV/2020]** [Learning to Learn with Variational Information Bottleneck for Domain Generalization](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123550205.pdf): meta-learning 11 | - **[ECCV/2020]** [Learning to Generate Novel Domains for Domain Generalization](https://arxiv.org/abs/2007.03304) 12 | - **[ECCV/2020]** [Learning to Optimize Domain Specific Normalization for Domain Generalization](https://arxiv.org/abs/1907.04275) 13 | - **[ECCV/2020]** [HGNet: Hybrid Generative Network for Zero-shot Domain Adaptation](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123720052.pdf) 14 | 15 | 16 | ### Medical related 17 | - **[IEEE-TMI/2020]** [MS-Net: Multi-Site Network for Improving Prostate Segmentation with Heterogeneous MRI Data](https://github.com/liuquande/MS-Net): **Supervised** domain specific BN. **Task**: Prostate Segmentation 18 | - **[Neurocomputing/2019]** [USE-Net: Incorporating Squeeze-and-Excitation blocks into U-Net for prostate zonal segmentation of multi-institutional MRI datasets](https://www.sciencedirect.com/science/article/pii/S0925231219309245): **Supervised** 19 | - **[MIDL-Abstract/2019]** [A Strong Baseline for Domain Adaptation and Generalization in Medical Imaging](https://openreview.net/forum?id=S1gvm2E-t4) 20 | - **[MICCAI/2018]** [Alifelong learning approach to brain mr segmentation across scannersand protocols](https://arxiv.org/pdf/1805.10170.pdf): DSBN. **Brain** structure segmentation in **MR** images 21 | 22 | ## Domain Adaptation (DA) 23 | 24 | ECCV2020 37 DA related papers 25 | - **[ECCV/2020]** [Instance Adaptive Self-Training for Unsupervised Domain Adaptation](http://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123710409.pdf): adaptation in the instance level 26 | - :star::star::star::star::star::star:[ECCV/2020/Oral]** [Domain-invariant Stereo Matching Newtorks](https://arxiv.org/pdf/1911.13287.pdf) Domain normalisation 27 | - **[ECCV/2020]** [Joint Disentangling and Adaptation for Cross-Domain Person Re-Identification](https://arxiv.org/abs/2007.10315) CycleGAN+disentanglement+self-training 28 | - **[ECCV/2020]** [Unsupervised Domain Adaptation with Noise Resistible Mutual-Training for Person Re-identification](https://zhaoj9014.github.io/pub/1391.pdf) 29 | - **[ECCV/2020]** [Unsupervised Domain Adaptation in the Dissimilarity Space for Person Re-identification](https://arxiv.org/abs/2007.13890) 30 | - **[ECCV/2020]** [Label Propagation with Augmented Anchors: A Simple Semi-Supervised Learning baseline for Unsupervised Domain Adaptation](https://arxiv.org/abs/2007.07695): Transfer DA to semi-supervised learning and performance is quite good. 31 | - **[ECCV/2020]** [Domain2Vec: Domain Embedding for Unsupervised Domain Adaptation](https://arxiv.org/abs/2007.09257): disentangle [domain + category] + 2 GAN **[CODE](https://github.com/VisionLearningGroup/Domain2Vec)** 32 | - **[ECCV/2020]** [Unsupervised Domain Attention Adaptation Network for Caricature Attribute Recognition](https://arxiv.org/abs/2007.09344) 33 | - **[ECCV/2020]** [Learning to Combine: Knowledge Aggregation for Multi-Source Domain Adaptation](https://arxiv.org/abs/2007.08801) 34 | - **[ECCV/2020]** [A Balanced and Uncertainty-aware Approach for Partial Domain Adaptation](https://arxiv.org/abs/2003.02541) 35 | - **[ECCV/2020]** [Class-Incremental Domain Adaptation](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123580052.pdf) source free Good 36 | - **[ECCV/2020]** [Spatial Attention Pyramid Network for Unsupervised Domain Adaptation](https://arxiv.org/pdf/2003.12979.pdf) 37 | - **[ECCV/2020]** [Attract, Perturb, and Explore: Learning a Feature Alignment Network for Semi-supervised Domain Adaptation](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123590579.pdf) 38 | - **[ECCV/2020]** [Curriculum Manager for Source Selection in Multi-Source Domain Adaptation](https://arxiv.org/pdf/2007.01261.pdf) 39 | - **[ECCV/2020]** [Two-phase Pseudo Label Densification for Self-training based Domain Adaptation](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123580528.pdf) 40 | - **[ECCV/2020]** [Attract, Perturb, and Explore: Learning a Feature Alignment Network for Semi-supervised Domain Adaptation](https://arxiv.org/abs/2007.09375) 41 | - **[ECCV/2020]** [Curriculum Manager for Source Selection in Multi-Source Domain Adaptation](https://arxiv.org/pdf/2007.01261.pdf): interesting. Simple method with curriculum learning 42 | - **[ECCV/2020]** [Learning to Detect Open Classes for Universal Domain Adaptation] 43 | - **[ECCV/2020]** [Contextual-Relation Consistent Domain Adaptation for Semantic Segmentation](https://arxiv.org/abs/2007.02424): local alignment 44 | - **[ECCV/2020]** [Partially-Shared Variational Auto-encoders for Unsupervised Domain Adaptation with Target Shift](https://arxiv.org/pdf/2001.07895.pdf) : interesting. Select informative instance to update networks. mutual learning with two networks. Like Yixiao Ge. 45 | - **[ECCV/2020]** [Online Meta-Learning for Multi-Source and Semi-Supervised Domain Adaptation](https://arxiv.org/abs/2004.04398) 46 | - **[ECCV/2020]** [Transferring Domain Shift Across Tasks for Zero-shot Domain adaptation] 47 | - **[ECCV/2020]** [YOLO in the Dark - Domain Adaptation Method for Merging Multiple Models] 48 | - **[ECCV/2020]** [Minimum Class Confusion for Versatile Domain Adaptation](https://arxiv.org/abs/1912.03699) 49 | - **[ECCV/2020]** [Learning from Scale-Invariant Examples for Domain Adaptation in Semantic Segmentation](https://arxiv.org/abs/2007.14449) 50 | - **[ECCV/2020]** [Task-conditioned Domain Adaptation for Pedestrian Detection in Thermal Imagery] 51 | - **[ECCV/2020]** [Universal Self-Training for Unsupervised Domain Adaptation] 52 | - **[ECCV/2020]** [Domain Adaptation through Task Distillation] 53 | - **[ECCV/2020]** [Multi-Source Open-Set Deep Adversarial Domain Adaptation](https://dipeshtamboli.github.io/blog/2020/Multi-Source-Open-Set-Deep-Adversarial-Domain-Adaptation/) 54 | - **[ECCV/2020]** [Label-Driven Reconstruction for Domain Adaptation in Semantic Segmentation](https://arxiv.org/abs/2003.04614) 55 | - **[ECCV/2020]** [High Resolution Zero-Shot Domain Adaptation of Synthetically Rendered Face Images](https://arxiv.org/abs/2006.15031) 56 | - **[ECCV/2020]** [Unsupervised Monocular Depth Estimation for Night-time Images using Adversarial Domain Feature Adaptation] 57 | - **[ECCV/2020]** [Dual Mixup Regularized Learning for Adversarial Domain Adaptation](https://arxiv.org/abs/2007.03141) 58 | - **[ECCV/2020]** [Skin Segmentation from NIR Images using Unsupervised Domain Adaptation through Generative Latent Search](https://arxiv.org/abs/2006.08696) 59 | 60 | 61 | ### Medical related 62 | 63 | - **[MICCAI/2020]** [Self Domain Adapted Network](https://link.springer.com/chapter/10.1007/978-3-030-59710-8_43): DA at test time on each image. OCT layer segmentation & MRI synthesis. using auto encoder between feature blocks. 64 | - **[TMI/2020]** [Anatomy-Regularized Representation Learning for Cross-Modality Medical Image Segmentation](https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9201096): An applicant ion of CV method. 65 | - **ISTN [MICCAI/2020]** [Image-level Harmonization of Multi-Site Datausing Image-and-Spatial Transformer Networks](https://arxiv.org/pdf/2006.16741.pdf): w/o comparison with other DA methods. ???? 66 | - **[CVPR/2020]** [Unsupervised Instance Segmentation in Microscopy Images via Panoptic DomainAdaptation and Task Re-weighting](http://openaccess.thecvf.com/content_CVPR_2020/papers/Liu_Unsupervised_Instance_Segmentation_in_Microscopy_Images_via_Panoptic_Domain_Adaptation_CVPR_2020_paper.pdf): CycleGAN + 3 kinds of (GRL) adaptation: image, instance, and semantic **Tasks**: nuclei segmentation of histopathology patch image 67 | - **[ICCV/2019]** [Semantic-Transferable Weakly-Supervised Endoscopic Lesions Segmentation](http://openaccess.thecvf.com/content_ICCV_2019/papers/Dong_Semantic-Transferable_Weakly-Supervised_Endoscopic_Lesions_Segmentation_ICCV_2019_paper.pdf): Endoscopic Lesions Segmentation. weakly supervised 68 | - **BEAL [MICCAI/2019]** [Boundary and Entropy-Driven Adversarial Learning for Fundus Image Segmentation](https://link.springer.com/chapter/10.1007/978-3-030-32239-7_12): boundary and entropy based adversairl learning. on **Fundus image segmentation** 69 | - **[MICCAI/2019]** [Unsupervised Domain Adaptation via Disentangled Representations: Application to Cross-Modality Liver Segmentation](https://link.springer.com/chapter/10.1007/978-3-030-32245-8_29). **Tasks**: CT-MR. **Liver** segmentation. 70 | - **SIFA [AAAI/2019]** [Synergistic image and feature adaptation: Towardscross-modality domain adaptation for medical image seg-mentation](https://arxiv.org/abs/1901.08211): CycleGAN variance. shared encoder for segfmenattion and image generation. **Tasks**: CT-MR 71 | - **[IEEE-TMI/2019]** [Patch-Based Output Space Adversarial Learning for Joint Optic Disc and Cup Segmentation](https://ieeexplore.ieee.org/abstract/document/8643416): output space adversarial learning. **Fundus image segmentation** 72 | - **SeUDA [MICCAI workshop/2018]** [Semantic-Aware Generative Adversarial Netsfor Unsupervised Domain Adaptationin Chest X-ray Segmentation](https://arxiv.org/pdf/1806.00600.pdf):CycleGAN **Tasks**: different chest X-ray 73 | - **[MICCAI/2018]** [Adversarial Domain Adaptation for Classification of Prostate Histopathology Whole-Slide Images](https://arxiv.org/abs/1806.01357):feature-level adverarial learning. **Task**: whole slide Prostate image classification 74 | - **[MICCAI/2018]** [Tumor-Aware, Adversarial DomainAdaptation from CT to MRI for LungCancer Segmentation](https://link.springer.com/content/pdf/10.1007%2F978-3-030-00934-2_86.pdf): CT<-->MRI. tumor segmentation --> semi-supervised. **Lung cancer** 75 | - **TD-GAN [MICCAI/2018]** [Task Driven Generative Modeling for Unsupervised Domain Adaptation: Application to X-ray Image Segmentation](https://link.springer.com/chapter/10.1007/978-3-030-00934-2_67):CycleGAN+Segmentation Net **Task**: CT-X-ray 76 | - **SeUDA [MICCAI-Workshop/2018]** [Semantic-aware gen-erative adversarial nets for unsupervised domain adaptation in chestx-ray segmentation](https://arxiv.org/abs/1806.00600) 77 | - **[MIDL/2018]** [Domain Adaptation for MRI Organ Segmentation using Reverse Classification Accuracy](https://arxiv.org/abs/1806.00363): supervised domain adaptation: two-center MR database, organ segmentation 78 | - **[arxiv/2018]** [Unsupervised domain adaptation for medical imaging segmentation with self-ensembling](https://arxiv.org/abs/1811.06042v1): multi-center data. Spinal Cord Gray Matter 79 | - **[JBHI/2017]** [Epithelium-Stroma Classification via Convolutional Neural Networks and Unsupervised Domain Adaptation in Histopathological Images ](https://ieeexplore.ieee.org/document/7893702/) 80 | - **[IPMI/2016]** [Unsupervised domain adaptation in brain lesionsegmentation with adversarial networks](https://arxiv.org/pdf/1612.08894.pdf): MR-->MR. **Brain** 81 | 82 | 83 | 84 | ### Domain randomization - 2D 85 | - **[CVPR/2020]** [Learning Texture Invariant Representationfor Domain Adaptation of Semantic Segmentation](http://openaccess.thecvf.com/content_CVPR_2020/papers/Kim_Learning_Texture_Invariant_Representation_for_Domain_Adaptation_of_Semantic_Segmentation_CVPR_2020_paper.pdf): like domain randomization, first generate styled images then train with usually GAN. **Datasets**: GTA5, SYNTHIA, Cityscapes 86 | - **[ICCV/2019]** []() 87 | 88 | ### Adversarial-based UDA - 2D Semantic segmentation 89 | - **[ICIP/2020]**[VARIATIONAL AUTOENCODER BASED UNSUPERVISED DOMAIN ADAPTATION FOR SEMANTIC SEGMENTATION])(https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9190973&tag=1) 90 | - **[AAAI/2020]** [Content-Consistent Matching for DomainAdaptive Semantic Segmentation](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123590426.pdf) layout matching 91 | - **[ECCV/2020]** [Self-Supervised CycleGAN for Object-Preserving Image-to-Image Domain Adaptation](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123650494.pdf): patch-based domain ailgnment based on CycleGAN+domain classificaion network 92 | - **[ECCV/2020]** [Spatial Attention Pyramid Network for Unsupervised Domain Adaptation](https://arxiv.org/pdf/2003.12979.pdf): Attention-based network architecture design. 93 | - **[ECCV/2020]** [Every Pixel Matters: Center-aware Feature Alignment for Domain Adaptive Object Detector](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123540698.pdf): Global feature GAN+class-specifc GAN 94 | - **[ECCV/2020]** [CSCL: Critical Semantic-Consistent Learning for Unsupervised Domain Adaptation](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123530732.pdf): RL-based GAN **[CODE](https://github.com/chengchunhsu/EveryPixelMatters)** 95 | - :star::star::star::star::star::star: **SIM [CVPR/2020]** [Differential Treatment for Stuff and Things:A Simple Unsupervised Domain Adaptation Method for Semantic Segmentation](http://openaccess.thecvf.com/content_CVPR_2020/papers/Wang_Differential_Treatment_for_Stuff_and_Things_A_Simple_Unsupervised_Domain_CVPR_2020_paper.pdf): pseudo label; Performance is better than oral paper below, solved GAN training decrease. **Datasets**: Sources[GTA5, SYNTHIA], Targ et[Cityscapes validation set] 96 | - **[CVPR/2020]** [Unsupervised Intra-domain Adaptation for Semantic Segmentationthrough Self-Supervision](https://arxiv.org/pdf/2004.07703v1.pdf): ADVENT variance, split the dataset into easy and hard subsets. Then adversaril learning between them by a discriminator. Improvement is evident compared with ADVENT but similar to MaxSquare. **Datasets**: Sources[GTA5, SYNTHIA, Synscapes], Targ et[Cityscapes validation set] 97 | - **[ECCV/2020]** [Two-phase Pseudo Label Densification for Self-training based Domain Adaptation](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123580528.pdf): similar to the above one. The performance even higher than above 98 | - **Dec [AAAI/2020]** [Joint Adversarial Learning for Domain Adaptationin Semantic Segmentation](https://www.aaai.org/Papers/AAAI/2020GB/AAAI-ZhangY.4858.pdf) **Datasets**:GTA5, Cityscapes, SYNTHIA 99 | - **[AAAI/2020]** [An Adversarial Perturbation Oriented Domain Adaptation Approach forSemantic Segmentation](https://arxiv.org/pdf/1912.08954v1.pdf) 100 | - :star::star::star::star::star::star:**[NIPS/2019]** [Category Anchor-Guided Unsupervised DomainAdaptation for Semantic Segmentation](https://arxiv.org/pdf/1910.13049.pdf) 101 | - **SIBAN [ICCV/2019]** [Significance-aware Information Bottleneck for Domain AdaptiveSemantic Segmentation](http://openaccess.thecvf.com/content_ICCV_2019/papers/Luo_Significance-Aware_Information_Bottleneck_for_Domain_Adaptive_Semantic_Segmentation_ICCV_2019_paper.pdf) **Datasets**: GTA5, Synthiam, Cityscapes. 102 | - **SSF-DAN [ICCV/2019]** [SSF-DAN: Separated Semantic Feature based Domain Adaptation Network forSemantic Segmentation](http://openaccess.thecvf.com/content_ICCV_2019/papers/Du_SSF-DAN_Separated_Semantic_Feature_Based_Domain_Adaptation_Network_for_Semantic_ICCV_2019_paper.pdf): Semantic-wise Separable Discriminator. **Datasets**: GTA5, Synthiam, Cityscapes. 103 | - **MaxSquare [ICCV/2019]** [Domain Adaptation for Semantic Segmentation with Maximum Squares Loss](http://openaccess.thecvf.com/content_ICCV_2019/papers/Chen_Domain_Adaptation_for_Semantic_Segmentation_With_Maximum_Squares_Loss_ICCV_2019_paper.pdf): Change loss function and entropy calculation way with square and margin with multi-scale training. **Datasets**: Office-31, Sources[GTA5, SYNTHIA], Target[Cityscapes validation set] 104 | - **Patch alignment [ICCV/2019]** [Domain Adaptation for Structured Output via DDiscriminative Patch Representations](https://arxiv.org/pdf/1901.05427.pdf): performance is not so good. **Datasets**: Sources[GTA5, SYNTHIA], Target[Cityscapes validation set] 105 | - **(TGCF-DA [ICCV/2019]** [Self-Ensembling with GAN-based Data Augmentation for Domain Adaptation inSemantic Segmentation](http://openaccess.thecvf.com/content_ICCV_2019/papers/Choi_Self-Ensembling_With_GAN-Based_Data_Augmentation_for_Domain_Adaptation_in_Semantic_ICCV_2019_paper.pdf): image transfer + mean teacher **Datasets**: GTA5, SYNTHIA, Cityscapes 106 | - **ADVENT [CVPR/2019]** [ADVENT: Adversarial Entropy Minimization for Domain Adaptationin Semantic Segmentation](http://openaccess.thecvf.com/content_CVPR_2019/papers/Vu_ADVENT_Adversarial_Entropy_Minimization_for_Domain_Adaptation_in_Semantic_Segmentation_CVPR_2019_paper.pdf): Then Adversarial learning based on entropy maps between source and target by a discriminator. 107 | - **CrDoCo [CVPR/2019]** [CrDoCo: Pixel-level Domain Transfer with Cross-Domain Consistency](http://openaccess.thecvf.com/content_CVPR_2019/papers/Chen_CrDoCo_Pixel-Level_Domain_Transfer_With_Cross-Domain_Consistency_CVPR_2019_paper.pdf): Cycle-GAN+2 segmentators. **Datasets**:GTA5, SYNTHIA, CITYSCAPES 108 | - **[CVPR/2019]** [Bidirectional Learning for Domain Adaptation of Semantic Segmentation](http://openaccess.thecvf.com/content_CVPR_2019/papers/Li_Bidirectional_Learning_for_Domain_Adaptation_of_Semantic_Segmentation_CVPR_2019_paper.pdf): Cycle-GAN+ sgementatio network train simutenously. **Datasets**:GTA5, SYNTHIA, CITYSCAPES 109 | - **[CVPR/2019]** [Not All Areas Are Equal: Transfer Learning for Semantic Segmentation viaHierarchical Region Selection](http://openaccess.thecvf.com/content_CVPR_2019/papers/Sun_Not_All_Areas_Are_Equal_Transfer_Learning_for_Semantic_Segmentation_CVPR_2019_paper.pdf) **Datasets**: GTAV + CITYSCAPES→CITYSCAPES, SYNTHIA +CITYSCAPES→CITYSCAPES and GTAV + SYNTHIA+ CITYSCAPES→CITYSCAPES. 110 | - **SPIGAN [ICLR/2019]** [SPIGAN: Privileged Adversarial Learning from Simulation](https://openreview.net/pdf?id=rkxoNnC5FQ) **Datasets**: SYNTHIA-RAND-CITYSCAPES. 111 | - **AdaptSegNet [CVPR/2018]** [Learning to Adapt Structured Output Space for Semantic Segmentation](http://openaccess.thecvf.com/content_cvpr_2018/papers/Tsai_Learning_to_Adapt_CVPR_2018_paper.pdf): multi-level output adversarial learning. **Datasets**: GTA5, SYNTHIA, Cross-City 112 | - **MCD-DA [CVPR/2018]** [Maximum Classifier Discrepancy for Unsupervised Domain Adaptation](http://openaccess.thecvf.com/content_cvpr_2018/papers/Saito_Maximum_Classifier_Discrepancy_CVPR_2018_paper.pdf): Not good on segmentation tasks. **Datasets**: classification and segmentation 113 | - **CyCADA [ICLR/2018]** [CyCADA: Cycle-Consistent Adversarial Domain Adaptation](http://proceedings.mlr.press/v80/hoffman18a/hoffman18a.pdf): Cycle-GAN + output Adversarial learning 114 | - **[ECCV/2018]** [DCAN: Dual Channel-wise Alignment Networksfor Unsupervised Scene Adaptation](https://eccv2018.org/openaccess/content_ECCV_2018/papers/Zuxuan_Wu_DCAN_Dual_Channel-wise_ECCV_2018_paper.pdf): CycleGAN+alignment based on channel mean and std. **Datasets**: Synthia, Gta5, Cityscapes. 115 | - **[Arxiv/2016]** [Fcns in the wild: Pixel-level adversarial and constraint-basedadaptation](https://arxiv.org/abs/1612.02649) **Datasets**: Cityscapes, SYNTHIA, GTA5, BDDS 116 | 117 | ### pseudo labels-based UDA - 2D Semantic segmentation 118 | - **[CVPR/2019]** [Domain-Specific Batch Normalization for Unsupervised Domain Adaptation](https://arxiv.org/pdf/1906.03950.pdf) 119 | - **[IJCAI/2020]** [Unsupervised Scene Adaptation with Memory Regularizationin vivo](https://arxiv.org/pdf/1912.11164.pdf): auxiliary classifier instead teacher model + pseudo label **Datasets**:GTA5, Cityscapes, SYNTHIA, Oxford RobotCar 120 | - **CBST [ECCV/2018]** [Unsupervised Domain Adaptation for SemanticSegmentation via Class-Balanced Self-Training](https://eccv2018.org/openaccess/content_ECCV_2018/papers/Yang_Zou_Unsupervised_Domain_Adaptation_ECCV_2018_paper.pdf): pseudo labels+class-balanced self-train **Datasets**:NTHU,Cityscapes, SYNTHIA, GTA5. 121 | - **BDL [CVPR/2019]** [Bidirectional Learning for Domain Adaptation of Semantic Segmentation](https://arxiv.org/pdf/1904.10620.pdf) 122 | 123 | ### Person Re-identification 124 | - **[ECCV/2020]** [Deep Credible Metric Learning for Unsupervised Domain Adaptation Person Re-identification](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123530630.pdf) 125 | - **[ECCV/2020]** [Generalizing Person Re-Identification by Camera-Aware Invariance Learning and Cross-Domain Mixup](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123600222.pdf): mixup variances 126 | - **[ICLR/2020]** [Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification](https://openreview.net/pdf?id=rJlnOhVYPS): dual mean teacher with contrastive learning. 127 | 128 | ### Other methods - 2D Semantic segmentation 129 | - **[ICCV/2019]** [Constructing Self-motivated Pyramid Curriculums for Cross-Domain SemanticSegmentation: A Non-Adversarial Approach](http://openaccess.thecvf.com/content_ICCV_2019/papers/Lian_Constructing_Self-Motivated_Pyramid_Curriculums_for_Cross-Domain_Semantic_Segmentation_A_Non-Adversarial_ICCV_2019_paper.pdf) **Datasets**: GTAV, SYNTHIA, Cityscapes. 130 | 131 | ### Semantic Nighttime Image Segmentation 132 | - **[ICCV/2019]** [Guided Curriculum Model Adaptation and Uncertainty-Aware Evaluation forSemantic Nighttime Image Segmentation](http://openaccess.thecvf.com/content_ICCV_2019/papers/Sakaridis_Guided_Curriculum_Model_Adaptation_and_Uncertainty-Aware_Evaluation_for_Semantic_Nighttime_ICCV_2019_paper.pdf) **Datasets**:Dark Zurich 133 | 134 | ### Domain invariant alignment - Object recognition 135 | - [**CVPR/2020**] [Structure Preserving Generative Cross-Domain Learning](https://openaccess.thecvf.com/content_CVPR_2020/papers/Xia_Structure_Preserving_Generative_Cross-Domain_Learning_CVPR_2020_paper.pdf): projection gragh matching 136 | - [Submitted to NIPS/2020] [Domain Adaptation without Source Data](https://arxiv.org/pdf/2007.01524.pdf) 137 | - **[ECCV/2020]** [Mind the Discriminability: Asymmetric Adversarial Domain Adaptation](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123690579.pdf): Auto encoder to reconstruct feature for S and T domains. 138 | - **[ECCV/2020]** [Class-Incremental Domain Adaptation](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123580052.pdf): project Source domain to target domain, align them 139 | - **[ICML/2020]** [Implicit Class-Conditioned Domain Alignment for Unsupervised Domain Adaptation](https://arxiv.org/abs/2006.04996). MMD+pseudo-labels. **Datasets**: Office-31, Office-Home, andVisDA2017. 140 | - **[CVPR/2020]** [Universal Source-Free Domain Adaptation](https://openaccess.thecvf.com/content_CVPR_2020/papers/Kundu_Universal_Source-Free_Domain_Adaptation_CVPR_2020_paper.pdf) 141 | - **DADA [AAAI/2020]** [Discriminative Adversarial Domain Adaptation](https://arxiv.org/pdf/1911.12036.pdf): multi-task [category and domainpredictions]. **Datasets**: Office-31, [Syn2Real](https://arxiv.org/abs/1806.09755) 142 | - **[ICML/2019]** [On Learning Invariant Representations for Domain Adaptation](http://proceedings.mlr.press/v97/zhao19a/zhao19a.pdf): Theoretic analysis. **Datasets**: digits 143 | - **BSP [ICML/2019]** [Transferability vs. Discriminability:Batch Spectral Penalization for Adversarial Domain Adaptation](http://proceedings.mlr.press/v97/chen19i/chen19i.pdf) **Datasets**: Office-31, Office-Home, VisDA-2017, Digits 144 | - **CADA [CVPR/2019]** [Attending to Discriminative Certainty for Domain Adaptation](http://openaccess.thecvf.com/content_CVPR_2019/papers/Kurmi_Attending_to_Discriminative_Certainty_for_Domain_Adaptation_CVPR_2019_paper.pdf) **Datasets**: Office-31, Office-Home, ImageCLEF 145 | - **TPN [CVPR/2019]** [Transferrable Prototypical Networks for Unsupervised Domain Adaptation](https://arxiv.org/pdf/1904.11227.pdf): class-level alignment. **Datasets**: digits, VisDA 2017, [GTA5, Synthia, and Cityscapes], VisDA 2018 146 | - **SWD [CVPR/2019]** [Sliced Wasserstein Discrepancy for Unsupervised Domain Adaptation](https://arxiv.org/pdf/1903.04064.pdf): class-level alignment. **Datasets**: digits, VisDA 2017 147 | - **TADA [AAAI/2019]** [Transferable Attention for Domain Adaptation](http://ise.thss.tsinghua.edu.cn/~mlong/doc/transferable-attention-aaai19.pdf): Cluster Alignment, similari to Dou. NIPS. **Datasets**: digits, Office-31, ImageCLEF-DA 148 | - **[AAAI/2019]** [Exploiting Local Feature Patterns for Unsupervised Domain Adaptation](https://arxiv.org/pdf/1811.05042.pdf) **Datasets**: Office-31, Office-Home 149 | - **SAFN [ICCV/2019]** [Larger Norm More Transferable: An Adaptive Feature Norm Approach forUnsupervised Domain Adaptation](http://openaccess.thecvf.com/content_ICCV_2019/papers/Xu_Larger_Norm_More_Transferable_An_Adaptive_Feature_Norm_Approach_for_ICCV_2019_paper.pdf) **Datasets**: VisDA2017,Office-Home, Office-31, ImageCLEF-DA. 150 | - **CAT [ICCV/2019]** [Cluster Alignment with a Teacher for Unsupervised Domain Adaptation](http://openaccess.thecvf.com/content_ICCV_2019/papers/Deng_Cluster_Alignment_With_a_Teacher_for_Unsupervised_Domain_Adaptation_ICCV_2019_paper.pdf) **Datasets**: Office-31, Office-Home 151 | - **ADR [ICLR/2018]** [Adversarial Dropout Regularization](https://arxiv.org/abs/1711.01575) Tasks: image classification and semantic segmentation. **Datasets**: digits, VisDA 2017, [GTA5, Synthia, and Cityscapes] 152 | - **MSTN [ICML/2018]** [Learning semantic representations for unsupervised domain adaptation](http://proceedings.mlr.press/v80/xie18c/xie18c.pdf): **Datasets**: Office-31, ImageCLEF-DA, digits 153 | - **[CVPR/2018]** [Maximum Classifier Discrepancy for Unsupervised Domain Adaptation](https://arxiv.org/pdf/1712.02560.pdf) **Datasets**: digits, VisDA 2017, [GTA5, Synthia, and Cityscapes] 154 | - **[CVPR/2018]** [Generate To Adapt: Aligning Domains using Generative Adversarial Networks](http://openaccess.thecvf.com/content_cvpr_2018/papers/Sankaranarayanan_Generate_to_Adapt_CVPR_2018_paper.pdf) **Datasets**: digits, Office-31, CAD synthetic datase, VISDA dataset 155 | - **[ECCV/2018]** [Deep Adversarial Attention Alignment forUnsupervised Domain Adaptation:the Benefit of Target Expectation Maximization](http://openaccess.thecvf.com/content_ECCV_2018/papers/Guoliang_Kang_Deep_Adversarial_Attention_ECCV_2018_paper.pdf) **Datasets**: digits, Office-31 156 | - **iCAN [CVPR/2018]** [Collaborative and Adversarial Network for Unsupervised domain adaptation](http://openaccess.thecvf.com/content_cvpr_2018/CameraReady/1410.pdf): multi-level GAN. **Datasets**: Office-31, ImageCLEF-DA 157 | - **SimNet [CVPR/2018]** [Unsupervised Domain Adaptation with Similarity Learning](http://openaccess.thecvf.com/content_cvpr_2018/papers/Pinheiro_Unsupervised_Domain_Adaptation_CVPR_2018_paper.pdf): similarity among C classes learning. **Datasets**: Digits,Office-31, VisDA 158 | - **[NIPS/2018]** [Conditional Adversarial Domain Adaptation](https://papers.nips.cc/paper/7436-conditional-adversarial-domain-adaptation.pdf) **Datasets**: Office-31, ImageCLEF-DA, Office-Home, Digits, [VisDA-2017](http://ai.bu.edu/visda-2017/) 159 | - **MADA [AAAI/2018]** [Multi-Adversarial Domain Adaptation](https://arxiv.org/pdf/1809.02176.pdf) **Datasets**: Office-31, ImageCLEF-DA 160 | - **ADDA [CVPR/2017]** [Adversarial Discriminative Domain Adaptation](https://arxiv.org/pdf/1702.05464.pdf) **Tasks**: Digits, NYU depth dataset 161 | - **JAN [ICML/2017]** [Deep Transfer Learning with Joint Adaptation Networks](https://arxiv.org/pdf/1605.06636.pdf): MMD. **Datasets**: Office-31, [ImageCLEF-DA](http://imageclef.org/2014/adaptation) 162 | - **RTN [NIPS/2016]** [Unsupervised Domain Adaptation with Residual Transfer Networks](https://arxiv.org/abs/1602.04433): MMD. **Datasets**: Office-31, Office-Caltech. 163 | - **DANN [JMLR/2016]** [Domain-adversarial training of neural networks](https://arxiv.org/pdf/1505.07818.pdf) **Tasks**: Image classification & Re-Identification 164 | - **DAN [ICML/2015]** [Learning Transferable Features with Deep Adaptation Network](https://arxiv.org/pdf/1502.02791.pdf): MMD. **Datasets**: Office-31, Office-10 + Caltech-10 165 | 166 | 167 | ### Clustering 168 | - **SRDC [CVPR/2020]** [Unsupervised Domain Adaptation via Regularized Conditional Alignment](http://openaccess.thecvf.com/content_CVPR_2020/papers/Tang_Unsupervised_Domain_Adaptation_via_Structurally_Regularized_Deep_Clustering_CVPR_2020_paper.pdf) **Datasets**: Office-31, ImageCLEF-DA, Office-Home 169 | - **CAN [CVPR/2019]** [Contrastive Adaptation Network for Unsupervised Domain Adaptation](http://openaccess.thecvf.com/content_CVPR_2019/papers/Kang_Contrastive_Adaptation_Network_for_Unsupervised_Domain_Adaptation_CVPR_2019_paper.pdf) **Datasets**: Office-31, VisDA-2017 170 | - **DWT-MEC [CVPR/2019]** [Unsupervised Domain Adaptation using Feature-Whitening and Consensus Loss](http://openaccess.thecvf.com/content_CVPR_2019/papers/Roy_Unsupervised_Domain_Adaptation_Using_Feature-Whitening_and_Consensus_Loss_CVPR_2019_paper.pdf) **Datasets**: MNIST↔USPS, MNIST↔SVHN,CIFAR-10↔STL,Office-Home 171 | - **SymNets [CVPR/2019]** [Domain-Symmetric Networks for Adversarial Domain Adaptation](https://arxiv.org/abs/1904.04663) **Datasets**: Office-31, ImageCLEF-DA, Office-Home. 172 | - **DIRT-T [ICLR/2018]** [A DIRT-T Approach to Unsupervised Domain Adaptation](https://openreview.net/pdf?id=H1q-TM-AW) **Datasets**: MNIST, MNIST-M,SVHN, SYN DIGITS, SYN SIGNS, GTSRB, CIFAR-10, and STL-10. 173 | 174 | 175 | 176 | 177 | ### 3D Semantic Segmentation/Point cloud 178 | - **[ECCV/2020]** [Monocular 3D Object Detection via Feature Domain Adaptation](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123540018.pdf) point cloud DA 179 | - **xMUDA [CVPR/2020]** [xMUDA: Cross-Modal Unsupervised Domain Adaptationfor 3D Semantic Segmentation](http://openaccess.thecvf.com/content_CVPR_2020/papers/Jaritz_xMUDA_Cross-Modal_Unsupervised_Domain_Adaptation_for_3D_Semantic_Segmentation_CVPR_2020_paper.pdf): Fusion 2D and 3D. 180 | 181 | ### Video domain adaptation 182 | - **[ECCV/2020]** [Shuffle and Attend: Video Domain Adaptation](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123570664.pdf): 183 | - [**ECCV/2020**] [Omni-sourced Webly-supervised Learning for Video Recognition](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123600664.pdf): 滑雪视频 MMLAB， 收集数据啥的 184 | - :star::star: :star:[**CVPR/2020**] [Action Segmentation with Joint Self-Supervised Temporal Domain Adaptation](https://openaccess.thecvf.com/content_CVPR_2020/papers/Chen_Action_Segmentation_With_Joint_Self-Supervised_Temporal_Domain_Adaptation_CVPR_2020_paper.pdf): GRL DA + frame permutation classifier 185 | - [**WACV/2020**] [Action Segmentation with Mixed Temporal Domain Adaptation](https://openaccess.thecvf.com/content_WACV_2020/papers/Chen_Action_Segmentation_with_Mixed_Temporal_Domain_Adaptation_WACV_2020_paper.pdf): local GRL DA+ global DA with domain attention 186 | - [**ICCV/2020**] [Temporal Attentive Alignment for Large-Scale Video Domain Adaptation](https://openaccess.thecvf.com/content_ICCV_2019/papers/Chen_Temporal_Attentive_Alignment_for_Large-Scale_Video_Domain_Adaptation_ICCV_2019_paper.pdf): domain attention 187 | - [**AAAI/2020**] [Generative Adversarial Networks for Video-to-Video Domain Adaptation](https://arxiv.org/abs/2004.08058): CycleGAN 188 | 189 | 190 | 191 | 192 | 193 | ## GAN Training 194 | 195 | 196 | 197 | - **[NIPS 2020]** [Differentiable Augmentation for Data-Efficient GAN Training](https://github.com/mit-han-lab/data-efficient-gans): Add Data augmentatio between G and D. results seem amazing. 198 | - 199 | 200 | 201 | 202 | 203 | 204 | ## Representation Learning 205 | 206 | - [] [Self-Supervised Ranking for Representation Learning](https://arxiv.org/pdf/2010.07258.pdf): 换了一种loss 207 | - SimCLR **[ICML/2020]** [A Simple Framework for Contrastive Learning of Visual Representations](https://github.com/google-research/simclr) 208 | - SimCLRv2 **[NIPS/2020]** [Big Self-Supervised Models are Strong Semi-Supervised Learners](https://github.com/google-research/simclr) 209 | - MOCO **[]** [Momentum contrast for unsupervised visual representation learning](https://arxiv.org/abs/1911.05722) 210 | - **C2L** **[MICCAI/2020]** [Comparing to Learn: Surpassing ImageNet Pretraining on Radiographs by Comparing Image Representations](https://arxiv.org/abs/2007.07423) a version of MOCO update, change the way to select positive and negative pairs. --------------------------------------------------------------------------------