生成對抗網路及其變體的論文彙總

黃小天發表於2017-04-21

生成對抗網路(GAN)是近段時間以來最受研究者關注的機器學習方法之一,深度學習泰斗 Yann LeCun 就曾多次談到 這種機器學習理念的巨大價值和未來前景。而各類 GAN 的變體也層出不窮,近日機器之心也報導過生成對抗網路的最新進展與論文集,而本文更注重於從 GAN 及其變體的角度對其論文做一個完整的梳理。

專案地址:https://deephunt.in/the-gan-zoo-79597dc8c347


每一週都會有關於 GAN 的新論文出現,你很難對其一一記錄,而眾多 GAN 的新命名又使其難上加難。如果你想了解更多關於 GAN 的資訊,可參閱 OpenAI 一篇有關生成模型的博文,或者 Goodfellow 於 NIPS 2016 所做的生成對抗網路主題演講


因此,下面是一個持續更新的最新列表,透過 GAN 名稱+論文(並附 arXiv 論文地址)的形式彙總並編排了所有出現的 GAN:


  • GAN—生成對抗網路(Generative Adversarial Networks):https://arxiv.org/abs/1406.2661

  • 3D-GAN—透過 3D 生成對抗網路建模學習機率性目標形潛在空間(Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling):https://arxiv.org/abs/1610.07584

  • AdaGAN—AdaGAN:增強生成模型(AdaGAN: Boosting Generative Models):http://arxiv.org/abs/1701.02386v1

  • AffGAN—影像超解析度的折舊 MAP 推斷(Amortised MAP Inference for Image Super-resolution):https://arxiv.org/abs/1610.04490

  • ALI—對抗性推斷學習(Adversarially Learned Inference):https://arxiv.org/abs/1606.00704

  • AMGAN—帶有最大化啟用標註資料的生成對抗網路(Generative Adversarial Nets with Labeled Data by Activation Maximization):http://arxiv.org/abs/1703.02000v1

  • AnoGAN—使用生成對抗模型的無監督異常檢測引導標記的發現(Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery):http://arxiv.org/abs/1703.05921v1

  • ArtGAN—ArtGAN: 使用條件範疇生成對抗網路合成藝術作品(ArtGAN: Artwork Synthesis with Conditional Categorial GANs):https://arxiv.org/abs/1702.03410

  • b-GAN—b-GAN: 生成對抗網路的統一架構(b-GAN: Unified Framework of Generative Adversarial Networks):https://openreview.net/pdf?id=S1JG13oee

  • Bayesian GAN—深度分層隱式模型(Deep and Hierarchical Implicit Models):https://arxiv.org/abs/1702.08896

  • BEGAN—BEGAN:邊界均衡生成對抗網路(BEGAN:Boundary Equilibrium Generative Adversarial Networks):http://arxiv.org/abs/1703.10717v2

  • BiGAN—對抗性特徵學習(Adversarial Feature Learning):http://arxiv.org/abs/1605.09782v7

  • BS-GAN—邊界查詢生成對抗網路(Boundary-Seeking Generative Adversarial Networks):http://arxiv.org/abs/1702.08431v1

  • CGAN—透過條件生成對抗網路實現多樣而自然的影像描述(Towards Diverse and Natural Image Descriptions via a Conditional GAN):http://arxiv.org/abs/1703.06029v1

  • CCGAN—語境條件性生成對抗網路的半監督學習(Semi-Supervised Learning with Context-Conditional Generative Adversarial Networks):https://arxiv.org/abs/1611.06430v1

  • CatGAN—類屬生成對抗網路的無監督和半監督學習(Unsupervised and Semi-supervised Learning with Categorical Generative Adversarial Networks):http://arxiv.org/abs/1511.06390v2

  • CoGAN—共軛生成對抗網路(Coupled Generative Adversarial Networks):http://arxiv.org/abs/1606.07536v2

  • Context-RNN-GAN—用於抽象推導圖表生成的語境性 RNN-GAN(Contextual RNN-GANs for Abstract Reasoning Diagram Generation):https://arxiv.org/abs/1609.09444

  • C-RNN-GAN—C-RNN-GAN:對抗訓練的連續性迴圈神經網路(C-RNN-GAN: Continuous recurrent neural networks with adversarial training):https://arxiv.org/abs/1611.09904

  • CVAE-GAN—CVAE-GAN: 透過非對稱訓練生成細密紋路的影像(Fine-Grained Image Generation through Asymmetric Training):https://arxiv.org/abs/1703.10155

  • CycleGAN—使用迴圈一致性對抗網路進行非成對圖到圖翻譯(Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks):https://arxiv.org/abs/1703.10593

  • DTN—無監督跨領域影像生成(Unsupervised Cross-Domain Image Generation):https://arxiv.org/abs/1611.02200

  • DCGAN—使用深度卷積生成對抗網路進行無監督表徵學習(Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks):https://arxiv.org/abs/1511.06434

  • DiscoGAN—使用生成對抗網路學習發現跨領域關係(Learning to Discover Cross-Domain Relations with Generative Adversarial Networks):http://arxiv.org/abs/1703.05192v1

  • DualGAN—DualGAN: 圖到圖翻譯的無監督對偶學習(Unsupervised Dual Learning for Image-to-Image Translation):http://arxiv.org/abs/1704.02510v1

  • EBGAN—基於能量的生成對抗網路(Energy-based Generative Adversarial Network):http://arxiv.org/abs/1609.03126v4

  • f-GAN—f-GAN:使用變分散度最小化訓練生成式神經取樣器(f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization):https://arxiv.org/abs/1606.00709

  • GoGAN—Gang of GANs: 使用最大間隔排序的生成對抗網路(Generative Adversarial Networks with Maximum Margin Ranking):https://arxiv.org/abs/1704.04865

  • GP-GAN—GP-GAN: 走近逼真的高解析度影像混合(Towards Realistic High-Resolution Image Blending):http://arxiv.org/abs/1703.07195v2

  • IAN—使用自省的對抗性網路進行神經影像編輯(Neural Photo Editing with Introspective Adversarial Networks):https://arxiv.org/abs/1609.07093

  • iGAN—在自然影像流形上的生成式視覺操作(Generative Visual Manipulation on the Natural Image Manifold):https://arxiv.org/abs/1609.03552v2

  • IcGAN—影像編輯的可逆條件生成對抗網路(Invertible Conditional GANs for image editing):https://arxiv.org/abs/1611.06355

  • ID-CGAN- 使用條件生成對抗網路的影像 De-raining(Image De-raining Using a Conditional Generative Adversarial Network):http://arxiv.org/abs/1701.05957v3

  • Improved GAN—生成對抗網路訓練的改進技術(Improved Techniques for Training GANs):https://arxiv.org/abs/1606.03498

  • InfoGAN—InfoGAN:資訊最大化生成對抗網路的可解釋性表徵學習(InfoGAN:Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets):http://arxiv.org/abs/1606.03657v1

  • LR-GAN—LR-GAN:用於影像生成的分層遞迴生成對抗網路(LR-GAN: Layered Recursive Generative Adversarial Networks for Image Generation):http://arxiv.org/abs/1703.01560v1

  • LSGAN—最小二乘生成對抗網路(Least Squares Generative Adversarial Networks):http://arxiv.org/abs/1611.04076v3

  • LS-GAN—利普希茨密度上的損失敏感型生成對抗網路(Loss-Sensitive Generative Adversarial Networks on Lipschitz Densities):http://arxiv.org/abs/1701.06264v5

  • MGAN—使用馬爾可夫過程的生成對抗網路預計算實時紋理合成(Precomputed Real-Time Texture Synthesis with Markovian Generative Adversarial Networks):https://arxiv.org/abs/1604.04382

  • MAGAN—MAGAN: 生成對抗網路的邊緣自適應(Margin Adaptation for Generative Adversarial Networks):http://arxiv.org/abs/1704.03817v1

  • MalGAN—基於生成對抗網路的黑箱攻擊的對抗性惡意例項生成(Generating Adversarial Malware Examples for Black-Box Attacks Based on GAN):http://arxiv.org/abs/1702.05983v1

  • MARTA-GAN—遙感影像的深度無監督表徵學習(Deep Unsupervised Representation Learning for Remote Sensing Images):https://arxiv.org/abs/1612.08879

  • McGAN—McGan: 均值和協方差特徵匹配生成對抗網路(Mean and Covariance Feature Matching GAN):http://arxiv.org/abs/1702.08398v1

  • MedGAN—使用生成對抗網路生成多標註的離散電子健康記錄(Generating Multi-label Discrete Electronic Health Records using Generative Adversarial Networks):http://arxiv.org/abs/1703.06490v1

  • MIX+GAN—生成對抗網路中的泛化與均衡(Generalization and Equilibrium in Generative Adversarial Nets /GANs):https://arxiv.org/abs/1703.00573v3

  • MPM-GAN—生成對抗網路多智慧體的資訊傳遞(Message Passing Multi-Agent GANs):https://arxiv.org/abs/1612.01294

  • MV-BiGAN—多視角生成對抗網路(Multi-view Generative Adversarial Networks):http://arxiv.org/abs/1611.02019v1

  • pix2pix—條件對抗網路的圖到圖翻譯(Image-to-Image Translation with Conditional Adversarial Networks):https://arxiv.org/abs/1611.07004

  • PPGN—即插即用生成網路:在潛在空間中生成條件迭代影像(Plug & Play Generative Networks: Conditional Iterative Generation of Images in Latent Space):https://arxiv.org/abs/1612.00005

  • PrGAN—從多物件 2D 視角歸納 3D 模型(3D Shape Induction from 2D Views of Multiple Objects):https://arxiv.org/abs/1612.05872

  • RenderGAN—RenderGAN:生成逼真標註資料(RenderGAN: Generating Realistic Labeled Data):https://arxiv.org/abs/1611.01331

  • RTT-GAN—可視段落生成的迴圈主題轉換 GAN(Recurrent Topic-Transition GAN for Visual Paragraph Generation):https://arxiv.org/abs/1703.07022v2

  • SGAN—堆疊 GAN(Stacked Generative Adversarial Networks):https://arxiv.org/abs/1612.04357v4

  • SGAN—空間 GAN 的紋理合成(Texture Synthesis with Spatial Generative Adversarial Networks):https://arxiv.org/abs/1611.08207

  • SAD-GAN—SAD-GAN:透過 GAN 合成自動駕駛(SAD-GAN: Synthetic Autonomous Driving using Generative Adversarial Networks):https://arxiv.org/abs/1611.08788v1

  • SalGAN—SalGAN:透過 GAN 預測視覺顯著度(SalGAN: Visual Saliency Prediction with Generative Adversarial Networks):https://arxiv.org/abs/1701.01081v2

  • SEGAN—SEGAN:語音增強 GAN(SEGAN: Speech Enhancement Generative Adversarial Network):https://arxiv.org/abs/1703.09452v1

  • SeqGAN—SeqGAN:具有策略梯度的序列 GAN ( SeqGAN: Sequence Generative Adversarial Nets with Policy Gradient):https://arxiv.org/abs/1609.05473v5

  • SketchGAN—用於草圖檢索的對抗訓練(Adversarial Training For Sketch Retrieval):https://arxiv.org/abs/1607.02748

  • SL-GAN—半隱 GAN:學習根據屬性生成和修改面部影像(Semi-Latent GAN: Learning to generate and modify facial images from attributes):https://arxiv.org/abs/1704.02166

  • SRGAN—使用一個 GAN 實現圖片逼真的單一影像超解析度(Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network):https://arxiv.org/abs/1609.04802v3

  • S²GAN—使用風格與結構對抗網路建模生成影像(Generative Image Modeling using Style and Structure Adversarial Networks):https://arxiv.org/abs/1603.05631v2

  • SSL-GAN—透過語境條件下的 GAN 實現半監督學習(Semi-Supervised Learning with Context-Conditional Generative Adversarial Networks):https://arxiv.org/abs/1611.06430v1

  • StackGAN—StackGAN:透過堆疊 GAN 合成文字到圖片的逼真影像(StackGAN: Text to Photo-realistic Image Synthesis with Stacked Generative Adversarial Networks):https://arxiv.org/abs/1612.03242v1

  • TGAN—時間 GAN(Temporal Generative Adversarial Nets):https://arxiv.org/abs/1611.06624v1

  • TAC-GAN—TAC-GAN—文字條件下的輔助生成器 GAN(TAC-GAN—Text Conditioned Auxiliary Classifier Generative Adversarial Network):https://arxiv.org/abs/1703.06412v2

  • TP-GAN—超越人臉旋轉:透過保有正面檢視合成打造用於逼真和身份的整體與區域性感知 GAN(Beyond Face Rotation: Global and Local Perception GAN for Photorealistic and Identity Preserving Frontal View Synthesis):https://arxiv.org/abs/1704.04086

  • Triple-GAN—三重 GAN(Triple Generative Adversarial Nets):https://arxiv.org/abs/1703.02291v2

  • VGAN—作為能量模型變分訓練的 GAN(Generative Adversarial Networks as Variational Training of Energy Based Models):https://arxiv.org/abs/1611.01799

  • VAE-GAN—使用學習的相似性度量進行超畫素自編碼(Autoencoding beyond pixels using a learned similarity metric):https://arxiv.org/abs/1512.09300

  • ViGAN—透過變分資訊 GAN 生成和編輯影像(Image Generation and Editing with Variational Info Generative AdversarialNetworks):https://arxiv.org/abs/1701.04568v1

  • WGAN—Wasserstein GAN:https://arxiv.org/abs/1701.07875v2

  • WGAN-GP—Wasserstein GAN 的改進訓練(Improved Training of Wasserstein GANs):https://arxiv.org/abs/1704.00028

  • WaterGAN—WaterGAN:實時校正單目水下影像色彩的無監督生成網路(WaterGAN: Unsupervised Generative Network to Enable Real-time Color Correction of Monocular Underwater Images):https://arxiv.org/abs/1702.07392v1


機器之心報導 GAN 相關文章


相關文章