Research map / Image generation / Text-to-image diffusion
Cross-Modal Contrastive Learning for Text-to-Image Generation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use XMC-GAN instead of OP-GAN | text-to-image generation · MS-COCO COCO-14 | FID: 24.70 -> 9.33 | Helped |
| use XMC-GAN instead of the compared recent models | text-to-image generation · MS-COCO COCO-14 | human preference for realism: 77.3% | Helped |
| use XMC-GAN instead of the compared recent models | text-to-image generation · MS-COCO COCO-14 | human preference for text alignment: 74.1% | Helped |
| use XMC-GAN instead of TReCS | text-to-image generation · LN-COCO | FID: 48.70 -> 14.12 | Helped |
| add sentence-image contrastive loss (ablation) | XMC-GAN · COCO-14 validation | FID: 39.28 -> 19.25 | Helped |
| add region-word contrastive loss (ablation) | XMC-GAN · COCO-14 validation | FID: 39.28 -> 24.38 | Helped |
| add discriminator-feature image-image contrastive loss (ablation) | XMC-GAN · COCO-14 validation | FID: 39.28 -> 29.71 | Helped |
| combine sentence-image and region-word contrastive losses (ablation) | XMC-GAN · COCO-14 validation | FID: 19.25 and 24.38 -> 14.25 | Helped |
| add VGG-feature image-image contrastive loss to sentence-image and region-word losses (ablation) | XMC-GAN · COCO-14 validation | FID: 14.25 -> 11.12 | Helped |
| replace self-modulation with attentional self-modulation (ablation) | XMC-GAN generator · COCO-14 validation | FID: 13.59 -> 11.93 | Helped |
| replace VGG L2 loss with VGG contrastive loss (ablation) | XMC-GAN · COCO-14 validation | FID: 52.86 -> 39.58 | Helped |
| add a 2-layer MLP discriminator projection head (ablation) | XMC-GAN · COCO-14 validation | FID: 12.61 -> 19.42 | Hurt |
Closest papers by meaning
- DF-GAN: A Simple and Effective Baseline for Text-to-Image Synthesis (CVPR 2022 · similarity 0.89)
- Adversarial Text to Continuous Image Generation (CVPR 2024 · similarity 0.86)
- E$^2$GAN: Efficient Training of Efficient GANs for Image-to-Image Translation (ICML 2024 · similarity 0.86)
- IT3D: Improved Text-to-3D Generation with Explicit View Synthesis (AAAI 2024 · similarity 0.85)
- DiffusionGAN3D: Boosting Text-guided 3D Generation and Domain Adaptation by Combining 3D GANs and Diffusion Priors (CVPR 2024 · similarity 0.85)
- InstantPainting: Expanding GANs for Efficient Text-Conditioned Image Generation Platform (AAAI 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.