New Study : https://www.notion.so/Reading-Papers-Deep-Learning-504b50ddaed14360b34dfd6d49cb3455

Update 2024.01.09

Paper Review

• 개인 공부라 열심히는 하고 있으나, 완벽한 리뷰가 아닙니다.
• 리뷰가 끝나더라도 계속 의문/생각/교정/좋은자료가 있다면 꾸준히 업데이트 됩니다.
• link review는 다른 분들이 하신 좋은 리뷰를 링크한 것입니다.
• light_link는 빠르게 개념(abstract)정도로 본 논문을 의미.
• 현재 상황이 리뷰 공개를 못하고 있는 상황이라, 논문 링크로만 정리진행합니다.

Virtual Try On [Link]

Asymmetric Image Retrieval [Link]

Diffusion [Link]

Deep Learning

• Revisiting Small Batch Training for Deep Neural Networks : [paper][review]
• Weight Standardization : [paper][link_review] [link_review]
• Effects of Image Size on Deep Learning : [paper]
• Inductive Bias : [link_review]

Multi-Label Image Recognition

• Learning Discriminative Representations for Multi-Label Image Recognition : [paper]

Knowledge distillation

• Knowledge distillation: A good teacher is patient and consistent : [paper]
• Hierarchical Self-supervised Augmented Knowledge Distillation : [paper]
• Text is Text, No Matter What: Unifying Text Recognition using Knowledge Distillation : [paper]

Vision and Language Pre-trained [Link]

CLIP & joined multi-modal

• CLIP : Learning Transferable Visual Models From Natural Language Supervision : [paper] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review]
• How Much Can CLIP Benefit Vision-and-Language Tasks? : [paper]
• Zero-Shot Open Set Detection by Extending CLIP : [paper]

Efficient training & trick

• Bag of Tricks for Image Classification with Convolutional Neural Networks : [paper] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review]

Imbalance Datasets

Self Supervised Learninig & unsupervised learning & semi/Weakly supervised learning

• Unsupervised Representation Learning by Predicting Image Rotations : [paper][]
• Unsupervised Visual Representation Learning by Context Prediction : [paper][]
• Unsupervised Learning of Visual Representations by Solving Jigsaw Puzzles : [paper][]
• Discriminative Unsupervised Feature Learning with Exemplar Convolutional Neural Networks : [paper][]
• Rethinking Pre-training and Self-training : [paper][]
• Selfie: Self-supervised Pretraining for Image Embedding : [paper] [light_review]
• Self-training with Noisy Student improves ImageNet classification : [paper] [review]
• SimCLR : A Simple Framework for Contrastive Learning of Visual Representations : [paper] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review]
• SimCLR V2:Big Self-Supervised Models are Strong Semi-Supervised Learners : [paper]
• MoCo : Momentum Contrast for Unsupervised Visual Representation Learning : [paper]
• MoCo V2 : Improved Baselines with Momentum Contrastive Learning : [paper] [link_review] [link_review]
• MoCo V3 : An Empirical Study of Training Self-Supervised Vision Transformers: [paper] [link_review] [link_review]
• BYOL : Bootstrap your own latent: A new approach to self-supervised Learning: [paper]
• Exploring the limits of weakly supervised pretraining : [paper]
• Triplet is All You Need with Random Mappings for Unsupervised Visual Representation Learning : [paper]
• ScatSimCLR: self-supervised contrastive learning with pretext task regularization for small-scale datasets : [paper]

Self Supervised Training + Mask based Token + Transformer

• MST: Masked Self-Supervised Transformer for Visual Representation : [paper]
• Masked Autoencoders Are Scalable Vision Learners : [paper]
• SimMIM: A Simple Framework for Masked Image Modeling : [paper]

Self Supervised Training + Instance Image Retrival

• InsCLR: Improving Instance Retrieval with Self-Supervision : [paper]

Vision Transformers classification

• Stand-Alone Self-Attention in Vision Models : [paper][review] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review]
• Selfie: Self-supervised Pretraining for Image Embedding : [paper] [light_review] [link_review] [link_review]
• ViT:An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale: [paper] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review]
• DeiT:Training data-efficient image transformers & distillation through attention : [paper] [link_review] [link_review] [link_review] [link_review]
• Bottleneck Transformers for Visual Recognition: [paper] [link_review]
• Going deeper with Image Transformers: [paper]
• Rethinking Spatial Dimensions of Vision Transformers : [paper]
• On the Adversarial Robustness of Visual Transformers: [paper]
• TransFG: A Transformer Architecture for Fine-grained Recognition : [paper]
• Understanding Robustness of Transformers for Image Classification : [paper]
• DeepViT: Towards Deeper Vision Transformer : [paper]
• CrossViT: Cross-Attention Multi-Scale Vision Transformer for Image Classification : [paper]
• CvT: Introducing Convolutions to Vision Transformers: [paper] [link_review]
• Efficient Feature Transformations for Discriminative and Generative Continual Learning : [paper]
• Swin Transformer: Hierarchical Vision Transformer using Shifted Windows : [paper] [link_review] [link_review] [link_review]
• Can Vision Transformers Learn without Natural Images?: [paper]
• Scaling Local Self-Attention for Parameter Efficient Visual Backbones: [paper]
• Incorporating Convolution Designs into Visual Transformers : [paper]
• ConViT: Improving Vision Transformers with Soft Convolutional Inductive Biases : [paper]
• Explicitly Modeled Attention Maps for Image Classification : [paper]
• Conditional Positional Encodings for Vision Transformers : [paper]
• Transformer in Transformer: [paper] [link_review]
• A Survey on Visual Transformer: [paper]
• Co-Scale Conv-Attentional Image Transformers: [paper]
• Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity : [paper] [link_review]
• LocalViT: Bringing Locality to Vision Transformers : [paper]
• Visformer: The Vision-friendly Transformer : [paper]
• Multiscale Vision Transformers : [paper] [link_review] [link_review]
• So-ViT: Mind Visual Tokens for Vision Transformer: [paper]
• Token Labeling: Training a 85.4% Top-1 Accuracy Vision Transformer with 56M Parameters on ImageNet (이후 "All Tokens Matter: Token Labeling for Training Better Vision Transformers 변경"): [paper]
• Fourier Image Transformer: [paper]
• Emerging Properties in Self-Supervised Vision Transformers: [paper]
• ConTNet: Why not use convolution and transformer at the same time?: [paper]
• Twins: Revisiting Spatial Attention Design in Vision Transformers: [paper]
• MoCo V3 :An Empirical Study of Training Self-Supervised Vision Transformers: [paper] [link_review] [link_review]
• Conformer: Local Features Coupling Global Representations for Visual Recognition: [paper]
• Self-Supervised Learning with Swin Transformers: [paper]
• Are Pre-trained Convolutions Better than Pre-trained Transformers?: [paper]
• LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference: [paper]
• Are Convolutional Neural Networks or Transformers more like human vision?: [paper]
• Rethinking Skip Connection with Layer Normalization in Transformers and ResNets: [paper]
• Rethinking the Design Principles of Robust Vision Transformer (Towards Robust Vision Transformer): [paper]
• Longformer: The Long-Document Transformer : [paper] [link_review] [link_review] [link_review] [link_review]
• Multi-Scale Vision Longformer: A New Vision Transformer for High-Resolution Image Encoding: [paper]
• On the Robustness of Vision Transformers to Adversarial Examples: [paper]
• Refiner: Refining Self-attention for Vision Transformers: [paper]
• Patch Slimming for Efficient Vision Transformers: [paper]
• RegionViT: Regional-to-Local Attention for Vision Transformers: [paper]
• X-volution: On the unification of convolution and self-attention: [paper]
• The Image Local Autoregressive Transformer: [paper]
• Glance-and-Gaze Vision Transformer: [paper]
• Semantic Correspondence with Transformers: [paper]
• DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification: [paper]
• When Vision Transformers Outperform ResNets without Pretraining or Strong Data Augmentations: [paper] [link_review]
• KVT: k-NN Attention for Boosting Vision Transformers: [paper]
• Less is More: Pay Less Attention in Vision Transformers: [paper]
• FoveaTer: Foveated Transformer for Image Classification: [paper]
• An Attention Free Transformer: [paper]
• Not All Images are Worth 16x16 Words: Dynamic Vision Transformers with Adaptive Sequence Length: [paper]
• Beyond Self-attention: External Attention using Two Linear Layers for Visual Tasks: [paper] [link_review]
• Pre-Trained Image Processing Transformer: [paper] [link_review]
• ResT: An Efficient Transformer for Visual Recognition: [paper]
• Towards Robust Vision Transformer: [paper]
• Aggregating Nested Transformers: [paper]
• GasHis-Transformer: A Multi-scale Visual Transformer Approach for Gastric Histopathology Image Classification: [paper]
• Intriguing Properties of Vision Transformers: [paper] [link_review] [link_review] [link_review]
• Vision Transformers are Robust Learners: [paper]
• Shuffle Transformer: Rethinking Spatial Shuffle for Vision Transformer: [paper]
• A Survey of Transformers: [paper]
• Armour: Generalizable Compact Self-Attention for Vision Transformers : [paper]
• Evo-ViT: Slow-Fast Token Evolution for Dynamic Vision Transformer : [paper]
• Dual-stream Network for Visual Recognition : [paper]
• BEiT: BERT Pre-Training of Image Transformers : [paper]
• Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolutions : [paper]
• PVTv2: Improved Baselines with Pyramid Vision Transformer : [paper]
• Thinking Like Transformers : [paper]
• CMT: Convolutional Neural Networks Meet Vision Transformers : [paper] [link_review] [link_review]
• Transformer with Peak Suppression and Knowledge Guidance for Fine-grained Image Recognition : [paper]
• ViTAE: Vision Transformer Advanced by Exploring Intrinsic Inductive Bias : [paper]
• Visual Transformer Pruning : [paper]
• Local-to-Global Self-Attention in Vision Transformers : [paper]
• Feature Fusion Vision Transformer for Fine-Grained Visual Categorization : [paper]
• Vision Xformers: Efficient Attention for Image Classification : [paper]
• EsViT : Efficient Self-supervised Vision Transformers for Representation Learning : [paper]
• GLiT: Neural Architecture Search for Global and Local Image Transformer : [paper]
• Efficient Vision Transformers via Fine-Grained Manifold Distillation : [paper]
• What Makes for Hierarchical Vision Transformer? : [paper]
• AutoFormer: Searching Transformers for Visual Recognition : [paper]
• Focal Self-attention for Local-Global Interactions in Vision Transformers : [paper] [link_review]
• ConvNets vs. Transformers: Whose Visual Representations are More Transferable? : [paper]
• Demystifying Local Vision Transformer: Sparse Connectivity, Weight Sharing, and Dynamic Weight : [paper]
• Mobile-Former: Bridging MobileNet and Transformer : [paper]
• Image Fusion Transformer : [paper]
• PSViT: Better Vision Transformer via Token Pooling and Attention Sharing : [paper]
• Do Vision Transformers See Like Convolutional Neural Networks? : [paper]
• Linformer: Self-Attention with Linear Complexity : [paper] [link_review]
• CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows : [paper]
• How to train your ViT? Data, Augmentation, and Regularization in Vision Transformers : [paper]
• Searching for Efficient Multi-Stage Vision Transformers : [paper]
• Exploring and Improving Mobile Level Vision Transformers : [paper]
• XCiT: Cross-Covariance Image Transformers : [paper]
• Efficient Vision Transformers via Fine-Grained Manifold Distillation : [paper]
• Scaled ReLU Matters for Training Vision Transformers : [paper]
• VOLO: Vision Outlooker for Visual Recognition : [paper]
• CoAtNet: Marrying Convolution and Attention for All Data Sizes : [paper] [link_review] [link_review]
• MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer : [paper]
• A free lunch from ViT: Adaptive Attention Multi-scale Fusion Transformer for Fine-grained Visual Recognition : [paper]
• Improved Multiscale Vision Transformers for Classification and Detection : [paper]

Vision Transformers positional embedding

• Self-Attention with Relative Position Representations : [paper] [link_review]
• Vision Transformer with Progressive Sampling : [paper]
• DPT: Deformable Patch-based Transformer for Visual Recognition : [paper]
• CAPE: Encoding Relative Positions with Continuous Augmented Positional Embeddings : [paper]
• Rethinking and Improving Relative Position Encoding for Vision Transformer : [paper]
• Rethinking Positional Encoding : [paper]
• Relative Positional Encoding for Transformers with Linear Complexity : [paper]
• Conditional Positional Encodings for Vision Transformers : [paper]
• Pyramid Adversarial Training Improves ViT Performance : [paper]
• Shunted Self-Attention via Multi-Scale Token Aggregation : [paper]
• AdaViT: Adaptive Vision Transformers for Efficient Image Recognition : [paper]
• ATS: Adaptive Token Sampling For Efficient Vision Transformers : [paper]
• Global Interaction Modelling in Vision Transformer via Super Tokens : [paper]

Vision Transformers vs MLP (or Others)

• AS-MLP: An Axial Shifted MLP Architecture for Vision : [paper]
• S2-MLPv2: Improved Spatial-Shift MLP Architecture for Vision : [paper]
• ResMLP: Feedforward networks for image classification with data-efficient training: [paper]
• Pay Attention to MLPs: [paper]
• Do You Even Need Attention? A Stack of Feed-Forward Layers Does Surprisingly Well on ImageNet: [paper]
• MLP-Mixer: An all-MLP Architecture for Vision : [paper]
• Sparse-MLP: A Fully-MLP Architecture with Conditional Computation : [paper]
• ConvMLP: Hierarchical Convolutional MLPs for Vision : [paper]
• Vision Permutator: A Permutable MLP-Like Architecture for Visual Recognition : [paper]
• MetaFormer is Actually What You Need for Vision : [paper]
• Adaptive Fourier Neural Operators: Efficient Token Mixers for Transformers : [paper] *

Vision Transformers retrieval

• Investigating the Vision Transformer Model for Image Retrieval Tasks: [paper]
• Training Vision Transformers for Image Retrieval: [paper]
• Instance-level Image Retrieval using Reranking Transformers: [paper]
• Retrieve Fast, Rerank Smart: Cooperative and Joint Approaches for Improved Cross-Modal Retrieval: [paper]
• TransHash: Transformer-based Hamming Hashing for Efficient Image Retrieval : [paper]
• Billion-Scale Pretraining with Vision Transformers for Multi-Task Visual Representations : [paper]
• Vision Transformer Hashing for Image Retrieval : [paper]

Vision Transformers segmentation and detection

• CoSformer: Detecting Co-Salient Object with Transformers: [paper]
• MDETR -- Modulated Detection for End-to-End Multi-Modal Understanding: [paper]
• Beyond Self-attention: External Attention using Two Linear Layers for Visual Tasks: [paper]
• Medical Image Segmentation Using Squeeze-and-Expansion Transformers: [paper]
• SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers: [paper]
• Visual Transformers: Token-based Image Representation and Processing for Computer Vision : [paper][]
• DETR:End-to-End Object Detection with Transformers : [paper] [link_review] [link_review] [link_review] [link_review] [link_review]
• Unifying Global-Local Representations in Salient Object Detection with Transformer : [paper]
• A Unified Efficient Pyramid Transformer for Semantic Segmentation : [paper]
• Dual-stream Network for Visual Recognition : [paper]
• MaX-DeepLab: End-to-End Panoptic Segmentation with Mask Transformers : [paper]
• Vision Transformers with Patch Diversification : [paper]
• Improve Vision Transformers Training by Suppressing Over-smoothing : [paper]
• SOTR: Segmenting Objects with Transformers : [paper]
• Simpler is Better: Few-shot Semantic Segmentation with Classifier Weight Transformer : [paper]
• Polyp-PVT: Polyp Segmentation with Pyramid Vision Transformers : [paper]
• Unifying Global-Local Representations in Salient Object Detection with Transformer : [paper]
• Conditional DETR for Fast Training Convergence : [paper]
• Fully Transformer Networks for Semantic Image Segmentation : [paper]
• Segmenter: Transformer for Semantic Segmentation : [paper]
• nnFormer: Interleaved Transformer for Volumetric Segmentation : [paper]
• Benchmarking Detection Transfer Learning with Vision Transformers : [paper]

Vision Transformers video

• An Image is Worth 16x16 Words, What is a Video Worth?: [paper]
• Token Shift Transformer for Video Classification : [paper]

Vision Transformers face

• Robust Facial Expression Recognition with Convolutional Visual Transformers : [paper]
• Learning Vision Transformer with Squeeze and Excitation for Facial Expression Recognition : [paper]

Vision Transformers OCR

• NRTR: A No-Recurrence Sequence-to-Sequence Model For Scene Text Recognition : [paper][]
• On Recognizing Texts of Arbitrary Shapes with 2D Self-Attention : [paper][]
• 2D Attentional Irregular Scene Text Recognizer : [paper][]

Vision Transformers multi-modal

• ReFormer: The Relational Transformer for Image Captioning : [paper]
• Long-Short Transformer: Efficient Transformers for Language and Vision : [paper]

Vision Transformers GAN

• A Hierarchical Transformation-Discriminating Generative Model for Few Shot Anomaly Detection : [paper]
• ViTGAN: Training GANs with Vision Transformers : [paper]
• Styleformer: Transformer based Generative Adversarial Networks with Style Vector : [paper]
• Combining Transformer Generators with Convolutional Discriminators : [paper]

Facebook AI Image Similarity Challenge

• 3rd Place: A Global and Local Dual Retrieval Solution to Facebook AI Image Similarity Challenge : [paper]

Google Landmark Challenge

Image Retrieval (Instance level Image Retrieval) & Deep Feature

• (My paper) All the attention you need: Global-local, spatial-channel attention for image retrieval : [paper]
• Large-Scale Image Retrieval with Attentive Deep Local Features : [paper] [review]
• NetVLAD: CNN architecture for weakly supervised place recognition : [paper][review]
• Learning visual similarity for product design with convolutional neural networks : [paper][review]
• Bags of Local Convolutional Features for Scalable Instance Search : [paper][review]
• Neural Codes for Image Retrieval : [paper][review]
• Conditional Similarity Networks : [paper][review]
• End-to-end Learning of Deep Visual Representations for Image Retrieval : [paper][review]
• CNN Image Retrieval Learns from BoW: Unsupervised Fine-Tuning with Hard Examples : [paper][review]
• Image similarity using Deep CNN and Curriculum Learning : [paper][review]
• Faster R-CNN Features for Instance Search : [paper][review]
• Regional Attention Based Deep Feature for Image Retrieval : [paper][review]
• Unsupervised Feature Learning via Non-Parametric Instance-level Discrimination : [paper][review]
• Object retrieval with deep convolutional features : [paper][review]
• Cross-dimensional Weighting for Aggregated Deep Convolutional Features : [paper][review]
• Learning Embeddings for Product Visual Search with Triplet Loss and Online Sampling : [paper][review]
• Saliency Weighted Convolutional Features for Instance Search : [paper][review]
• 2018 Google Landmark Retrieval Challenge 리뷰 : [review]
• 2019 Google Landmark Retrieval Challenge 리뷰 : [review]
• REMAP: Multi-layer entropy-guided pooling of dense CNN features for image retrieval : [paper][review]
• Large-scale Landmark Retrieval/Recognition under a Noisy and Diverse Dataset : [paper][review]
• Fine-tuning CNN Image Retrieval with No Human Annotation : [paper][review]
• Large Scale Landmark Recognition via Deep Metric Learning : [paper][review]
• Deep Aggregation of Regional Convolutional Activations for Content Based Image Retrieval : [paper][review]
• Challenging deep image descriptors for retrieval in heterogeneous iconographic collections : [paper][review]
• A Benchmark on Tricks for Large-scale Image Retrieval : [paper][review]
• Attention-Aware Generalized Mean Pooling for Image Retrieval : [paper][review]
• Class-Weighted Convolutional Features for Image Retrieval : [paper][review] # 100th
• deep image retrieval loss (계속 업데이트):[paper][review]
• Matchable Image Retrieval by Learning from Surface Reconstruction:[paper][review]
• Combination of Multiple Global Descriptors for Image Retrieval:[paper][review]
• Unifying Deep Local and Global Features for Efficient Image Search:[paper][review]
• ACTNET: end-to-end learning of feature activations and multi-stream aggregation for effective instance image retrieval:[paper][review]
• Google Landmarks Dataset v2 A Large-Scale Benchmark for Instance-Level Recognition and Retrieval:[paper][review]
• Detect-to-Retrieve: Efficient Regional Aggregation for Image Search:[paper][review]
• Local Features and Visual Words Emerge in Activations:[paper][review]
• Image Retrieval using Multi-scale CNN Features Pooling: [paper][review]
• MultiGrain: a unified image embedding for classes and instances: [paper][link_review] [link_review]
• Divide and Conquer the Embedding Space for Metric Learning: [paper][link_review]
• An Effective Pipeline for a Real-world Clothes Retrieval System: [paper][light_review]
• Instance Similarity Learning for Unsupervised Feature Representation : [paper]
• Towards Accurate Localization by Instance Search : [paper]
• The 2021 Image Similarity Dataset and Challenge : [paper]
• DOLG:Single-Stage Image Retrieval with Deep Orthogonal Fusion of Local and Global Features : [paper]
• Towards A Fairer Landmark Recognition Dataset : [paper]
• Recall@k Surrogate Loss with Large Batches and Similarity Mixup : [paper]

Metric Learning

• Deep metric learning using Triplet network : [paper][review]
• FaceNet: A Unified Embedding for Face Recognition and Clustering : [paper][review]
• Sampling Matters in Deep Embedding Learning : [paper][review]
• Online Progressive Deep Metric Learning : [paper]

Fashion Image Retrieval

• Learning Embeddings for Product Visual Search with Triplet Loss and Online Sampling : [paper][review]
• Conditional Similarity Networks : [paper][review]
• Semi-supervised Feature-Level Attribute Manipulation for Fashion Image Retrieval : [paper][link_review]

Fashion Compatibility & Outfit Recommendation

• Context-Aware Visual Compatibility Prediction: [paper][review] [light_review]
• Learning Type-Aware Embeddings for Fashion Compatibility : [paper] [review]
• OutfitNet: Fashion Outfit Recommendation with Attention-Based Multiple Instance Learning : [paper]

Personalized Outfit Recommendation & fashion outfit

• FashionNet: Personalized Outfit Recommendation with Deep Neural Network: [paper][review]
• Self-supervised Visual Attribute Learning for Fashion Compatibility : [paper]
• Personalized Outfit Recommendation with Learnable Anchors : [paper]
• PAI-BPR: Personalized Outfit Recommendation Scheme with Attribute-wise Interpretability : [paper]
• Hierarchical Fashion Graph Network for Personalized Outfit Recommendation : [paper]

Fashion multi-modal

• Kaleido-BERT: Vision-Language Pre-training on Fashion Domain : [paper]
• Fashion IQ: A New Dataset Towards Retrieving Images by Natural Language Feedback : [paper]

Fashion DataSets

• SHIFT15M: Multiobjective Large-Scale Fashion Dataset with Distributional Shifts : [paper]

Retail & Product & Instance

• Product1M: Towards Weakly Supervised Instance-Level Product Retrieval via Cross-modal Pretrainingr : [paper]
• RP2K: A Large-Scale Retail Product Dataset for Fine-Grained Image Classification : [paper]
• eProduct: A Million-Scale Visual Search Benchmark to Address Product Recognition Challenges : [paper]
• Regional Maximum Activations of Convolutions with Attention for Cross-domain Beauty and Personal Care Product Retrieval:[paper][review]
• Learning visual similarity for product design with convolutional neural networks : [paper][review]
• Product1M: Towards Weakly Supervised Instance-Level Product Retrieval via Cross-modal Pretrainingr : [paper]
• he Met Dataset: Instance-level Recognition for Artworks : [paper]

Image Retrieval using Deep Hash

• Deep Learning of Binary Hash Codes for Fast Image Retrieval : [paper][review]
• Feature Learning based Deep Supervised Hashing with Pairwise Labels : [paper][review]
• Deep Supervised Hashing with Triplet Labels : [paper][review]
• Online Hashing with Similarity Learning : [paper]

Video Classification

• NetVLAD: CNN architecture for weakly supervised place recognition : [paper][review]
• Learnable pooling with Context Gating for video classification : [paper][review]
• Less is More: Learning Highlight Detection from Video Duration : [paper][review]
• Efficient Video Classification Using Fewer Frames : [paper][review]

OCR - Recognition

• Synthetically Supervised Feature Learning for Scene Text Recognition : [paper][review]
• FOTS: Fast Oriented Text Spotting with a Unified Network : [paper][review]
• Robust Scene Text Recognition with Automatic Rectification : [paper][review]
• Joint Visual Semantic Reasoning: Multi-Stage Decoder for Text Recognition : [paper]

OCR - Detection

• PixelLink: Detecting Scene Text via Instance Segmentation : [paper][review]
• EAST: An Efficient and Accurate Scene Text Detector : [paper][review]
• Scene Text Detection with Supervised Pyramid Context Network : [paper][review]
• FOTS: Fast Oriented Text Spotting with a Unified Network : [paper][review]
• Character Region Awareness for Text Detection : [paper][review]

Attention & Deformation

• Squeeze Excitation Networks : [paper][review]
• Spatial Transformer Network : [paper][review]
• Tell Me Where to Look: Guided Attention Inference Network : [paper][review]
• CBAM: Convolutional Block Attention Module : [paper][review]
• BAM: Bottleneck Attention Module : [paper][review]
• Neural Machine Translation by Jointly Learning to Align and Translate : [paper][review]
• Residual Attention Networks for Image Classification : [paper][review]
• Attention is all you need : [paper][review][link_review]
• Residual Attention Network for Image Classification : [paper][review]
• Stand-Alone Self-Attention in Vision Models : [paper][review] [light_review] [light_review] [light_review] [light_review] [light_review]

Visual & Textual Embedding

• DeViSE: A Deep Visual-Semantic Embedding Model : [paper][review]
• Dual Attention Networks for Multimodal Reasoning and Matching : [paper][review]
• Learning Deep Structure-Preserving Image-Text Embeddings : [paper][review]
• Learning Two-Branch Neural Networks for Image-Text Matching Tasks : [paper] [link_review]

CNN

• Imagenet classification with deep convolutional neural networks : [paper][review]
• Going Deeper with Convolutions : [paper][review]
• ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices : [paper][review]
• Deep Residual Learning for Image Recognition : [paper][review]
• Aggregated Residual Transformations for Deep Neural Networks : [paper][review]
• Very Deep Convolutional Networks for Large-Scale Image Recognition : [paper][review]
• Squeeze Excitation Networks : [paper][review]
• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications : [paper][review]
• Pelee: A Real-Time Object Detection System on Mobile Devices : [paper][review]
• Residual Attention Network for Image Classification : [paper][review]
• Wide Residual Networks : [paper][review]
• Stand-Alone Self-Attention in Vision Models : [paper][review]
• Selective Kernel Networks : [paper][review]
• EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks : [paper] [link_review]
• CSPNet: A New Backbone that can Enhance Learning Capability of CNN : [paper] [link_review] [link_review]
• RepMLP: Re-parameterizing Convolutions into Fully-connected Layers for Image Recognition : [paper]

Transfer Learning

• Taskonomy: Disentangling Task Transfer Learning : [paper][link_review]
• What makes ImageNet good for transfer learning? : [paper][review]

Generative Adversarial Nets

• Generative Adversarial Nets : [paper][review]
• Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks : [paper][review]
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks : [paper][review]
• Progressive Growing of GANs for Improved Quality, Stability, and Variation : [paper][review]
• Beholder-GAN: Generation and Beautification of Facial Images with Conditioning on Their Beauty Level : [paper][review]
• Synthetically Supervised Feature Learning for Scene Text Recognition : [paper][review]
• A Style-Based Generator Architecture for Generative Adversarial Networks : [paper][review]
• High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs : [paper][review]
• Everybody Dance Now : [paper][review]
• Be Your Own Prada: Fashion Synthesis with Structural Coherence : [paper][review]
• Fashion-Gen: The Generative Fashion Dataset and Challenge : [paper][review]
• StackGAN++: Realistic Image Synthesis with Stacked Generative Adversarial Networks : [paper][review]
• DwNet: Dense warp-based network for pose-guided human video generation: [paper][review]

Face

• FaceNet: A Unified Embedding for Face Recognition and Clustering : [paper][review]
• The Devil of Face Recognition is in the Noise : [paper][link_review]
• Revisiting a single-stage method for face detection : [paper][review
• MixFaceNets: Extremely Efficient Face Recognition Networks : [paper]

Pose Estimation

• Convolutional Pose Machines : [paper][review]

NLP/NLU

• Efficient Estimation of Word Representations in Vector Space : [paper][review]
• node2vec: Scalable Feature Learning for Networks : [paper][review]
• Transfomer(self attention) 기본 이해하기 : PPT정리
• BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding : [paper][review](~ing)

Learning to Rank

• DeepRank: A New Deep Architecture for Relevance Ranking in Information Retrieval : [paper][review]
• SNRM: From Neural Re-Ranking to Neural Ranking: Learning a Sparse Representation for Inverted Indexing : [paper][review]
• TF-Ranking: Scalable TensorFlow Library for Learning-to-Rank : [paper][review]
• ConvRankNet: Deep Neural Network for Learning to Rank Query-Text Pairs : [paper][review]
• KNRM: End-to-End Neural Ad-hoc Ranking with Kernel Pooling : [paper][review]
• Conv-KNRM: Convolutional Neural Networks for Soft-Matching N-Grams in Ad-hoc Search : [paper][review]
• PACRR: A position-aware neural IR model for relevance matching : [paper][link_review]
• CEDR: Contextualized Embeddings for Document Ranking #262 : [paper][link]
• Deeper Text Understanding for IR with Contextual Neural Language Modeling : [paper][link]
• Simple Applications of BERT for Ad Hoc Document Retrieval : [paper][link]
• Document Expansion by Query Prediction : [paper][link]
• Passage Re-ranking with BERT : [paper][link]

Domain Adaptation

• Domain-Adversarial Training of Neural Networks : [paper][review]

Curriculum Learning

• CurriculumNet: Weakly Supervised Learning from Large-Scale Web Images : [paper][review]

Image Segmentation

• U-Net: Convolutional Networks for Biomedical Image Segmentation : [paper][review]
• Mask R-CNN : [paper][review]
• Fully Convolutional Networks for Semantic Segmentation : [paper][review]
• Cascade Decoder: A Universal Decoding Method for Biomedical Image Segmentation : [paper] [review]
• FickleNet: Weakly and Semi-supervised Semantic Image Segmentation using Stochastic Inference : [paper] [link_review] [link_review]
• Path Aggregation Network for Instance Segmentation : [paper] [link_review]

Localization

• YOLO: Real-Time Object Detection : [paper][review]
• YOLO9000: Better, Faster, Stronger : [paper][review]
• YOLOv4: Optimal Speed and Accuracy of Object Detection : [paper] [link_review] [link_review] [link_review] [link_review] [link_review] [link_review]
• Scaled-YOLOv4: Scaling Cross Stage Partial Network : [paper]
• Faster R-CNN : [paper][review]
  • faster rcnn의 anchor generator 개념 뿐만 아니라 소스레벨에서도 이해하기 : [review]
• SSD: Single Shot MultiBox Detector : [paper][link_review]
  • Why normalization performed only for conv4_3? : [review]
• Pelee: A Real-Time Object Detection System on Mobile Devices : [paper][review]
• R-FCN: Object Detection via Region-based Fully Convolutional Networks: [paper][review]
• Revisiting a single-stage method for face detection: [paper][review]
• DSSD : Deconvolutional Single Shot Detector: [paper][review]
• Feature-fused SSD: fast detection for small objects : [paper][link_review]
• EfficientDet ： Scalable and Efficient Object Detection : [paper] [link_review] [review]
• FCOS: Fully Convolutional One-Stage Object Detection : [paper] [light_review]
• Bridging the Gap Between Anchor-based and Anchor-free Detection via Adaptive Training Sample Selection : [paper] [light_review]
• Oriented R-CNN for Object Detection : [paper]
• CSPNet: A New Backbone that can Enhance Learning Capability of CNN : [paper] [link_review] [link_review]

AutoML

• Learning Transferable Architectures for Scalable Image Recognition : [paper][link_review]

Image Quality

• Learning to Compose with Professional Photographs on the Web : [paper][review]
• Photo Aesthetics Ranking Network with Attributes and Content Adaptation : [paper][review]
• Composition-preserving Deep Photo Aesthetics Assessment : [paper][review]
• Deep Image Aesthetics Classification using Inception Modules and Fine-tuning Connected Layer : [paper][review]
• NIMA: Neural Image Assessment : [paper][review]

Others

• Neural Arithmetic Logic Units : [paper][link_review]