Summary of deep learning papers

Summarize some interesting paper about computer vision

Image Classification Methods

:sunny: (LeNet)[Gradient-based learning applied to document recognition] [Paper]

:sunny: (AlexNet)[ImageNet Classification with Deep Convolutional Neural Networks] [NIPS 2012][Paper][Code]

:sunny: (VGGNet)[Very Deep Convolutional Networks for Large-Scale Image Recognition] [arXiv][Paper][Code]

:sunny: (GoogLeNet)[Going deeper with convolutions] [CVPR 2015][Paper][Code]

:sunny: (ResNet)[Deep Residual Learning for Image Recognition] [CVPR 2016][Paper][Code]

:sunny: (ResNeXt)[Aggregated Residual Transformations for Deep Neural Networks] [CVPR 2017][Paper][Code]

:sunny: (DenseNet)[Densely Connected Convolutional Networks] [CVPR 2017][Paper][Code]

:sunny: (Inception-v4)[Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning] [AAAI 2017][Paper][Code]

:sunny: (Inception-v3)[Rethinking the Inception Architecture for Computer Vision] [CVPR 2016][Paper][Code]

:sunny: (Xception)[Xception: Deep Learning with Depthwise Separable Convolutions] [CVPR 2017][Paper][Code]

:sunny: (ShuffleNet)[ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices] [arXiv][Paper][Code]

:sunny: (MobileNets)[MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications] [arXiv][Paper][Code]

:sunny: (SENet)[Squeeze-and-Excitation Networks] [CVPR 2018][Paper][Code]

Normalization Methods

:moon: (Batch Normalization)[Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift] [ICML][Paper]

:moon: (Instance Normalization)[Instance Normalization: The Missing Ingredient for Fast Stylization] [arXiv][Paper][Code]

:moon: (Layer Normalization)[Layer Normalization] [arXiv][Paper]

:moon: (Group Normalization)[Group Normalization] [arXiv][Paper][Code]

:moon: (Switchable Normalization)[Differentiable Learning-to-Normalize via Switchable Normalization] [arXiv][Paper][Code]

:moon: (Instance-Batch Normalization)[Two at Once: Enhancing Learning and Generalization Capacities via IBN-Net] [arXiv][Paper][Code]

Activation Functions

:star: (ReLU)[Rectified linear units improve restricted boltzmann machines] [ICML][Paper]

:star: (Noisy ReLU)[Deep Belief Networks on CIFAR-10] [arXiv][Paper]

:star: (Leaky ReLU)[Rectifier Nonlinearities Improve Neural Network Acoustic Models] [ICML 2013][Paper]

:star: (eLU)[Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)] [arXiv][Paper]

:star: (SeLU)[Self-Normalizing Neural Networks] [NIPS 2017][Paper][Code]

:star: (PReLU)[Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification] [ICCV 2015][Paper]

:star: (Maxout)[Maxout Networks] [JMLR 2013][Paper]

:star: (Swish)[Searching for Activation Functions] [arXiv][Paper]

Generative Adversarial Networks Theory

:blossom: (GAN)[Generative adversarial nets] [NIPS 2014][Paper][Code]

:blossom: (cGAN)[Conditional Generative Adversarial Nets] [arXiv][Paper][Code]

:blossom: (DCGAN)[Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks] [arXiv][Paper][Code]

:blossom: (LAPGAN)[Deep generative image models using a Laplacian pyramid of adversarial networks] [NIPS 2015][Paper][Code]

:blossom: (Semi-Supervised GAN)[Improved Techniques for Training GANs] [NIPS 2016][Paper][Code]

:blossom: (Info GAN)[InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets] [NIPS 2016][Paper][Code]

:blossom: (LSGAN)[Least Squares Generative Adversarial Networks] [ICCV 2017][Paper][Code]

:blossom: (WGAN)[Wasserstein Generative Adversarial Networks] [ICML 2017][Paper][Code]

:blossom: (WGAN-GP)[Improved Training of Wasserstein GANs] [NIPS 2017][Paper][Code]

:blossom: (EBGAN)[Energy-based Generative Adversarial Network] [arXiv][Paper][Code]

:blossom: (BEGAN)[BEGAN: Boundary Equilibrium Generative Adversarial Networks] [arXiv][Paper][Code]

:blossom: (PG-GAN)[Progressive Growing of GANs for Improved Quality, Stability, and Variation] [arXiv][Paper][Code]

:blossom: (SNGAN)[Spectral Normalization for Generative Adversarial Networks] [arXiv][Paper][Code]

:blossom: (DRAGAN)[On Convergence and Stability of GANs] [arXiv][Paper][Code]

:blossom: (Relativistic GAN)[The relativistic discriminator: a key element missing from standard GAN] [arXiv][Paper][Code]

:blossom: (cGAN with projection disc)[cGANs with projection discriminator] [arXiv][Paper][Code]

:blossom: (BigGAN)[Large Scale GAN Training for High Fidelity Natural Image Synthesis] [arXiv][Paper][Code]

:blossom: (SAGAN)[Self-Attention Generative Adversarial Networks] [arXiv][Paper][Code]

Image-to-Image Translation

:bug: (pix2pix)[Image-to-Image Translation with Conditional Adversarial Networks] [CVPR 2017][Paper][Code]

:bug: (pix2pixHD)[High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs] [CVPR 2018][Paper][Code]

:bug: (CycleGAN)[Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks] [ICCV 2017][Paper][Code]

:bug: [Unsupervised Attention-guided Image to Image Translation] [arXiv][Paper][Code]

:bug: (DiscoGAN)[Learning to Discover Cross-Domain Relations with Generative Adversarial Networks] [ICML 2017][Paper][Code]

:bug: (UNIT)[Unsupervised Image-to-Image Translation Networks] [NIPS 2017][Paper][Code]

:bug: (MUNIT)[Multimodal Unsupervised Image-to-Image Translation] [arXiv][Paper][Code]

:bug: (BicycleGAN)[Toward Multimodal Image-to-Image Translation] [NIPS 2017][Paper][Code]

:bug: (StarGAN)[StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation] [CVPR 2018][Paper][Code]

:bug: (RecycleGAN)[Recycle-GAN: Unsupervised Video Retargeting] [ECCV 2018][Paper][Code]

Style Transfer

:whale: (Gatys)[A Neural Algorithm of Artistic Style] [Paper]

:whale: (Johnson)[Perceptual Losses for Real-Time Style Transfer and Super-Resolution] [Paper]

:whale: [Universal Style Transfer via Feature Transforms] [Paper]

:whale: [Arbitrary Style Transfer in Real-time with Adaptive Instance Normalization] [Paper]

:whale: [Visual attribute transfer through deep image analogy] [Paper]

:whale: [Arbitrary Style Transfer with Deep Feature Reshuffle] [Paper]

:whale: [Artistic style transfer for videos] [Paper]

:whale: [Characterizing and Improving Stability in Neural Style Transfer] [Paper]

:whale: [Controlling Perceptual Factors in Neural Style Transfer] [Paper]

:whale: [Deep Photo Style Transfer] [Paper]

:whale: [Fast Patch-based Style Transfer of Arbitrary Style] [Paper]

:whale: [Improved Texture Networks: Maximizing Quality and Diversity in Feed-forward Stylization and Texture Synthesis] [Paper]

:whale: [Multimodal Transfer: A Hierarchical Deep Convolutional Neural Network for Fast Artistic Style Transfer] [Paper]

:whale: [StyleBank: An Explicit Representation for Neural Image Style Transfer] [Paper]

:whale: [CartoonGAN: generative adversarial networks for photo cartoonization] [Paper]

:whale: [Visual Attribute Transfer through Deep Image Analogy] [Paper]

:whale: [A learned representation for artistic style] [Paper]

Face Attribute Manipulation

:alien: [Deep Feature Interpolation for Image Content Changes] [Paper]

:alien: [Autoencoding beyond pixels using a learned similarity metric] [Paper]

:alien: [Convolutional Network for Attribute-driven and Identity-preserving Human Face Generation] [Paper]

:alien: [Deep Identity-aware Transfer of Facial Attributes] [Paper]

:alien: [Learning Residual Images for Face Attribute Manipulation] [Paper]

:alien: (StarGAN)[StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation] [Paper]

Image Inpainting

:eyes: [Context Encoders: Feature Learning by Inpainting] [Paper]

:eyes: [Semantic Image Inpainting with Deep Generative Models] [Paper]

:eyes: [High-Resolution Image Inpainting using Multi-Scale Neural Patch Synthesis] [Paper]

:eyes: [Globally and locally consistent image completion] [Paper]

:eyes: [Generative Image Inpainting with Contextual Attention] [Paper]

Super Resolution

:ox: [Image Super-Resolution Using Deep Convolutional Networks] [Paper]

:ox: [Enhanced Deep Residual Networks for Single Image Super-Resolution] [Paper]

:ox: [Residual Dense Network for Image Super-Resolution] [Paper]

To be continued.