CoDA-Gen: Contrastive-Optimized Diffusion Adaptation for Efficient Few-Shot Image Generation
Abstract
Despite their success in generating high-quality images, diffusion models encounter significant challenges in data-scarce environments such as few-shot image generation (FSIG). This limitation often leads to issues like mode collapse, poor visual quality, and insufficient diversity. To address these critical problems, we propose CoDA-Gen, a novel framework for Contrastive-Optimized Diffusion Adaptation for Generative Models. CoDA-Gen integrates a lightweight, contrastiveoptimized (CoDA) module into a pre-trained diffusion U-Net, enabling efficient learning and adaptation to the unique feature distributions of a target domain from extremely few samples. Our method leverages a contrastive learning objective to align generated features with those of few-shot target samples, thereby mitigating mode collapse and enhancing both the quality and diversity of generated images. Extensive experiments on MiniImageNet, CIFAR-FS, and CelebA-HQ demonstrate that CoDA-Gen consistently outperforms state-of-the-art baselines, achieving superior FID, IS, LPIPS, and Recall scores. Furthermore, CoDA-Gen exhibits remarkable adaptation efficiency, converging rapidly and highlighting its practical utility for rapid domain specialization.
Read the paper
Related studies
- Learning from First-Generation Qualitative Approaches in the IS Discipline: An Evolutionary View and Some Implications for Authors and Evaluators (PART 1/2)
- Commentary on the clinical and preclinical dosage limits of interstitially administered magnetic fluids for therapeutic hyperthermia based on current practice and efficacy models
- Prosthesis: New Technological Opportunities and Innovative Biomedical Devices
- The history and future of neural modeling for cochlear implants
- Ryanodine receptors: waking up from refractoriness
- Novel Transgenic Rabbit Model Sheds Light on the Puzzling Role of Matrix Metalloproteinase-12 in Atherosclerosis
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.