Research map / Image generation / Text-to-image diffusion
TISE: Bag of Metrics for Text-to-Image Synthesis Evaluation
Tan M. Dinh, Rang Nguyen, Binh-Son Hua
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 spectral normalization to the AttnGAN discriminators (AttnGAN++) (ablation) | AttnGAN · CUB | IS / FID / R-precision: IS 4.32 -> 4.78, FID 24.27 -> 15.01, RP 65.30 -> 77.31 | Improved |
| calibrate the Inception classifier with temperature scaling (IS to IS*) (ablation) | Inception Score · CUB | IS vs IS*: Counter Model IS 5.12 vs AttnGAN++ 4.78; IS* 13.05 vs 15.13 | Improved |
| calibrate the ImageNet Inception-v3 classifier (T=0.909) for IS* (ablation) | Inception Score on Tiny ImageNet · Tiny ImageNet | IS ranking: T = 0.909 | No clear effect |
| use CLIP encoders instead of DAMSM encoders for R-precision (ablation) | R-precision · MS-COCO | R-precision on real images: Real Images 67.35 -> 83.65; AttnGAN++ 96.39 -> 73.37 | Improved |
| compute SOA with Mask-RCNN instead of YOLO-v3 (ablation) | SOA · MS-COCO | SOA-C / SOA-I: CPGAN SOA-C 77.02 -> 82.25 versus Real Images 74.97 -> 89.98 | Improved |
| add contrastive learning (CL) to DM-GAN | DM-GAN · CUB | FID / R-precision: FID 15.52 -> 14.57, RP 76.25 -> 69.80 | Mixed |
| add contrastive learning (CL) to DM-GAN | DM-GAN · MS-COCO | Ranking score: 41.0 -> 51.5 | Improved |
| add contrastive learning (CL) to AttnGAN | AttnGAN · MS-COCO | Ranking score: 29.0 -> 37.0 | Improved |
| include IS* in the ranking | Ranking score · MS-COCO | IS* vs real images: IS* AttnGAN++ 54.63, CPGAN 59.64 vs Real Images 51.25 | Hurt |
| AttnGAN++ ranking score vs human evaluation (RS aligned with human score) | Ranking score · MS-COCO human study | Human score: AttnGAN 37.40 -> AttnGAN++ 45.01; RS 13.5 -> 28.5 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- AttnGAN — AttnGAN: Fine-Grained Text to Image Generation With Attentional Generative Adversarial Networks (CVPR 2018) · better on 5 of 5 shared results
- DM-GAN — DM-GAN: Dynamic Memory Generative Adversarial Networks for Text-To-Image Synthesis (CVPR 2019) · better on 4 of 4 shared results
- DF-GAN — DF-GAN: A Simple and Effective Baseline for Text-to-Image Synthesis (CVPR 2022) · better on 0 of 1 shared result
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)
- Revisiting text-to-image evaluation with Gecko: on metrics, prompts, and human rating (ICLR 2025 · similarity 0.86)
- Who Evaluates the Evaluations? Objectively Scoring Text-to-Image Prompt Coherence Metrics with T2IScoreScore (TS2) (NeurIPS 2024 · similarity 0.86)
- Q-Eval-100K: Evaluating Visual Quality and Alignment Level for Text-to-Vision Content (CVPR 2025 · similarity 0.86)
- ImagenHub: Standardizing the evaluation of conditional image generation models (ICLR 2024 · similarity 0.85)
- Rethinking FID: Towards a Better Evaluation Metric for Image Generation (CVPR 2024 · similarity 0.85)
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.
