Research map / Image generation / Text-to-image diffusion
Trace Controlled Text to Image Generation
Kun Yan, Lei Ji, Chenfei Wu, Jianmin Bao, Ming Zhou, Nan Duan, Shuai Ma
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| add mouse trace input (ablation) | TCTIG text-only · LN-COCO val (GAN synthesis ablation) | SSCP: 0.727 -> 0.923 | Improved |
| add mouse trace input (ablation) | TCTIG text-only · LN-COCO val (GAN synthesis ablation) | FID: 75.34 -> 29.58 | Improved |
| add segment aligned augmentation (SAA) (ablation) | TCTIG (-TGR) · LN-COCO val (GAN synthesis ablation) | FID: 29.58 -> 26.50 | Improved |
| add segment aligned augmentation (SAA) (ablation) | TCTIG (-TGR) · LN-COCO val (GAN synthesis ablation) | SOA-I: 0.1320 -> 0.2200 | Improved |
| add segment aligned augmentation (SAA) (ablation) | TCTIG (-TGR) · LN-COCO val (GAN synthesis ablation) | SSCP: 0.923 -> 0.955 | Improved |
| add trace guided re-weighting loss (TGR) (ablation) | TCTIG · LN-COCO val (GAN synthesis ablation) | FID: 26.50 -> 19.54 | Improved |
| add trace guided re-weighting loss (TGR) (ablation) | TCTIG · LN-COCO val (GAN synthesis ablation) | SSCP: 0.955 -> 0.962 | Improved |
| replace GAN-based VQGAN decoder with VQ-conditioned diffusion decoder (ablation) | TCTIG · LN-COCO val (GAN synthesis ablation) | FID: 19.54 -> 11.94 | Improved |
| replace GAN-based VQGAN decoder with VQ-conditioned diffusion decoder (ablation) | TCTIG · LN-COCO val (GAN synthesis ablation) | SSCP: 0.962 -> 0.973 | Improved |
| replace GAN decoder with diffusion decoder on ground truth tokens (ablation) | GT image reconstruction · LN-COCO val (GAN synthesis ablation) | FID: 12.74 -> 8.62 | Improved |
| use TCTIG instead of TRECS | trace-controlled generation · LN-COCO val | SSCP: 0.862 -> 0.973 | Improved |
| use TCTIG instead of TRECS | trace-controlled generation · LN-COCO val | SOA-I: 0.3523 -> 0.2658 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- AttnGAN — AttnGAN: Fine-Grained Text to Image Generation With Attentional Generative Adversarial Networks (CVPR 2018) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation (NeurIPS 2025 · similarity 0.87)
- LLM Blueprint: Enabling Text-to-Image Generation with Complex and Detailed Prompts (ICLR 2024 · similarity 0.87)
- Text-to-Image Generation for Abstract Concepts (AAAI 2024 · similarity 0.87)
- Brush Your Text: Synthesize Any Scene Text on Images via Diffusion Model (AAAI 2024 · similarity 0.86)
- ConceptMix: A Compositional Image Generation Benchmark with Controllable Difficulty (NeurIPS 2024 · similarity 0.86)
- Elucidating the design space of language models for image generation (ICML 2025 · similarity 0.86)
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.
