MaCo-GAN: Manifold-Contrastive Adversarial Learning for Single Image Super-Resolution
2026-06-03 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors address a problem in image enhancement called Single Image Super-Resolution, where current AI models sometimes create unrealistic details. Instead of judging the whole image, their method, MaCo-GAN, focuses on comparing realistic and fake image features in a more precise way. They generate tricky fake images based on real ones and train the AI to tell apart subtle differences, improving the balance between making images clear and natural-looking. Their approach works better without changing much of the existing model design.
Generative Adversarial Networks (GANs)Single Image Super-Resolution (SISR)adversarial losscontrastive learningmanifoldperception-distortion trade-offfake sample synthesizerdiscriminatorgeneratorconditional realism
Authors
Daeyoung Han, Seongmin Hwang, Moongu Jeon
Abstract
Conventional Generative Adversarial Networks (GANs) for Single Image Super-Resolution (SISR) often struggle with hallucinated artifacts, largely because standard discriminators evaluate overall image naturalness rather than strict conditional realism. To address this, we propose MaCo-GAN, a novel manifold-contrastive GAN framework that replaces the conventional adversarial loss with a supervised contrastive objective. A core component of our method is a dynamic fake sample synthesizer that transforms ground truth (GT) data into a spectrum of challenging, perceptually plausible fake images that strictly maintain low-resolution (LR) correspondence. Utilizing these synthesized samples, we establish a robust contrastive minimax game: the generator is trained to attract its predictions toward on-manifold fakes (low distortion) and repel them from off-manifold fakes (high distortion), while the discriminator optimizes the exact opposite. By simply replacing the adversarial loss of a baseline SR model with our proposed objective, we demonstrate consistent improvements in the perception-distortion trade-off across various benchmarks. Extensive ablation studies validate the effectiveness of our framework and provide deep insights into the dynamics of this conditional contrastive game.