TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Papers/Triple Generative Adversarial Networks

Triple Generative Adversarial Networks

Chongxuan Li, Kun Xu, Jiashuo Liu, Jun Zhu, Bo Zhang

2019-12-20Data AugmentationGeneral ClassificationImage GenerationClassificationConditional Image GenerationSemi-Supervised Image Classification
PaperPDFCode(official)

Abstract

We propose a unified game-theoretical framework to perform classification and conditional image generation given limited supervision. It is formulated as a three-player minimax game consisting of a generator, a classifier and a discriminator, and therefore is referred to as Triple Generative Adversarial Network (Triple-GAN). The generator and the classifier characterize the conditional distributions between images and labels to perform conditional generation and classification, respectively. The discriminator solely focuses on identifying fake image-label pairs. Under a nonparametric assumption, we prove the unique equilibrium of the game is that the distributions characterized by the generator and the classifier converge to the data distribution. As a byproduct of the three-player mechanism, Triple-GAN is flexible to incorporate different semi-supervised classifiers and GAN architectures. We evaluate Triple-GAN in two challenging settings, namely, semi-supervised learning and the extreme low data regime. In both settings, Triple-GAN can achieve excellent classification results and generate meaningful samples in a specific class simultaneously. In particular, using a commonly adopted 13-layer CNN classifier, Triple-GAN outperforms extensive semi-supervised learning methods substantially on more than 10 benchmarks no matter data augmentation is applied or not.

Results

TaskDatasetMetricValueModel
Image ClassificationCIFAR-10, 4000 LabelsPercentage error6.54Triple-GAN-V2 (ResNet-26)
Image ClassificationCIFAR-10, 4000 LabelsPercentage error10.01Triple-GAN-V2 (CNN-13)
Image ClassificationCIFAR-10, 4000 LabelsPercentage error12.41Triple-GAN-V2 (CNN-13, no aug)
Image ClassificationSVHN, 500 LabelsAccuracy96.39Triple-GAN-V2 (CNN-13)
Image ClassificationSVHN, 500 LabelsAccuracy96.16Triple-GAN-V2 (CNN-13, no aug)
Image ClassificationCIFAR-10, 1000 LabelsAccuracy91.59Triple-GAN-V2 (ResNet-26)
Image ClassificationCIFAR-10, 1000 LabelsAccuracy85Triple-GAN-V2 (CNN-13)
Image ClassificationCIFAR-10, 1000 LabelsAccuracy81.81Triple-GAN-V2 (CNN-13, no aug)
Image ClassificationSVHN, 1000 labelsAccuracy96.55Triple-GAN-V2 (CNN-13)
Image ClassificationSVHN, 1000 labelsAccuracy96.04Triple-GAN-V2 (CNN-13, no aug)
Image ClassificationSVHN, 250 LabelsAccuracy96.52Triple-GAN-V2 (CNN-13)
Image ClassificationSVHN, 250 LabelsAccuracy95.81Triple-GAN-V2 (CNN-13, no aug)
Semi-Supervised Image ClassificationCIFAR-10, 4000 LabelsPercentage error6.54Triple-GAN-V2 (ResNet-26)
Semi-Supervised Image ClassificationCIFAR-10, 4000 LabelsPercentage error10.01Triple-GAN-V2 (CNN-13)
Semi-Supervised Image ClassificationCIFAR-10, 4000 LabelsPercentage error12.41Triple-GAN-V2 (CNN-13, no aug)
Semi-Supervised Image ClassificationSVHN, 500 LabelsAccuracy96.39Triple-GAN-V2 (CNN-13)
Semi-Supervised Image ClassificationSVHN, 500 LabelsAccuracy96.16Triple-GAN-V2 (CNN-13, no aug)
Semi-Supervised Image ClassificationCIFAR-10, 1000 LabelsAccuracy91.59Triple-GAN-V2 (ResNet-26)
Semi-Supervised Image ClassificationCIFAR-10, 1000 LabelsAccuracy85Triple-GAN-V2 (CNN-13)
Semi-Supervised Image ClassificationCIFAR-10, 1000 LabelsAccuracy81.81Triple-GAN-V2 (CNN-13, no aug)
Semi-Supervised Image ClassificationSVHN, 1000 labelsAccuracy96.55Triple-GAN-V2 (CNN-13)
Semi-Supervised Image ClassificationSVHN, 1000 labelsAccuracy96.04Triple-GAN-V2 (CNN-13, no aug)
Semi-Supervised Image ClassificationSVHN, 250 LabelsAccuracy96.52Triple-GAN-V2 (CNN-13)
Semi-Supervised Image ClassificationSVHN, 250 LabelsAccuracy95.81Triple-GAN-V2 (CNN-13, no aug)

Related Papers

Overview of the TalentCLEF 2025: Skill and Job Title Intelligence for Human Capital Management2025-07-17Pixel Perfect MegaMed: A Megapixel-Scale Vision-Language Foundation Model for Generating High Resolution Medical Images2025-07-17fastWDM3D: Fast and Accurate 3D Healthy Tissue Inpainting2025-07-17Synthesizing Reality: Leveraging the Generative AI-Powered Platform Midjourney for Construction Worker Detection2025-07-17FashionPose: Text to Pose to Relight Image Generation for Personalized Fashion Visualization2025-07-17A Distributed Generative AI Approach for Heterogeneous Multi-Domain Environments under Data Sharing constraints2025-07-17Adversarial attacks to image classification systems using evolutionary algorithms2025-07-17Similarity-Guided Diffusion for Contrastive Sequential Recommendation2025-07-16