Research map / Image generation / Text-to-image diffusion
Semantic Object Accuracy for Generative Text-to-Image Synthesis
Tobias Hinz, Stefan Heinrich, Stefan Wermter
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 object pathways to generator and all discriminators | AttnGAN · MS-COCO | SOA-C: 25.88 -> 34.51 | Improved |
| add object pathways to generator and all discriminators | AttnGAN · MS-COCO | FID: 33.10 -> 24.70 | Improved |
| use object pathway at test time instead of 0 objects (ablation) | OPv2 + BBL + MO · MS-COCO | SOA-C: 16.55 -> 34.51 | Improved |
| use object pathway at test time instead of 0 objects (ablation) | OPv2 + BBL + MO · MS-COCO | FID: 45.79 -> 25.80 | Improved |
| add bounding box loss (ablation) | OPv2 · MS-COCO | SOA-C: 33.82 -> 33.19 | No clear effect |
| add bounding box loss (ablation) | OPv2 · MS-COCO | SOA-IoU-C: 0.207 -> 0.210 | Improved |
| train with up to ten objects per image (ablation) | OPv2 · MS-COCO | SOA-C: 33.82 -> 33.46 | No clear effect |
| train with up to ten objects per image (ablation) | OPv2 · MS-COCO | FID: 26.52 -> 24.94 | Improved |
| train with up to ten objects per image (ablation) | OPv2 · MS-COCO | R-precision: 87.73 -> 89.14 | Improved |
| combine bounding box loss and many objects (ablation) | OPv2 · MS-COCO | SOA-C: 33.82 -> 34.51 | Improved |
| apply object pathway to ten objects instead of three (ablation) | OPv2 trained with three objects · MS-COCO | SOA-C | No clear effect |
| OP-GAN versus DM-GAN on SOA | OP-GAN · MS-COCO | SOA-C: 33.44 -> 35.85 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DM-GAN — DM-GAN: Dynamic Memory Generative Adversarial Networks for Text-To-Image Synthesis (CVPR 2019) · better on 3 of 5 shared results
- AttnGAN — AttnGAN: Fine-Grained Text to Image Generation With Attentional Generative Adversarial Networks (CVPR 2018) · better on 5 of 5 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
- Evaluating Numerical Reasoning in Text-to-Image Models (NeurIPS 2024 · similarity 0.87)
- OneIG-Bench: Omni-dimensional Nuanced Evaluation for Image Generation (NeurIPS 2025 · similarity 0.87)
- Leveraging Unpaired Data for Vision-Language Generative Models via Cycle Consistency (ICLR 2024 · similarity 0.87)
- FineGRAIN: Evaluating Failure Modes of Text-to-Image Models with Vision Language Model Judges (NeurIPS 2025 · similarity 0.87)
- Elucidating the design space of language models for image generation (ICML 2025 · similarity 0.86)
- ImagenHub: Standardizing the evaluation of conditional image generation models (ICLR 2024 · 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.
