Summary of A Framework For Image Synthesis Using Supervised Contrastive Learning, by Yibin Liu et al.
A Framework For Image Synthesis Using Supervised Contrastive Learning
by Yibin Liu, Jianyu Zhang, Li Zhang, Shijian Li, Gang Pan
First submitted to arxiv on: 5 Dec 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper proposes a novel framework for text-to-image (T2I) generation using Generative Adversarial Network (GANs). Traditional T2I GANs are two-phase methods that pretrain an inter-modal representation and then train an image generator. However, these models ignore the inner-modal semantic correspondence between images with similar labels. The proposed framework leverages both inter- and inner-modal correspondence by label-guided supervised contrastive learning, using two parameter-sharing contrast branches in both pretraining and generation phases. This integration fosters the generation of higher-quality images by clustering semantically similar image-text pair representations. Experimental results on four novel T2I GANs demonstrate significant improvements in Inception Score (IS) and Frechet Inception Distance (FID) metrics, with notable improvements on the complex multi-object COCO dataset. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making pictures from text descriptions. Right now, computers can make some good pictures but not always accurate ones. To improve this, researchers propose a new way to teach computers to generate images based on texts. They use something called Generative Adversarial Network (GAN) and add an extra step to help the computer understand the meaning of similar images. The results show that this new method works better than previous methods in making accurate pictures. |
Keywords
» Artificial intelligence » Clustering » Gan » Generative adversarial network » Pretraining » Supervised
