Research map / Image generation / Autoregressive and latent generation
Hierarchical Image Generation via Transformer-Based Sequential Patch Selection
Xiaogang Xu, Ning Xu
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 |
|---|---|---|---|
| Use the proposed semi-parametric image generation framework | Scene-graph-to-image generation · COCO-Stuff, predicted boxes | FID: 56.9 -> 51.6 | Improved |
| Use the proposed semi-parametric image generation framework | Scene-graph-to-image generation · Visual Genome, predicted boxes | FID: 107.0 -> 63.7 | Improved |
| Use the proposed semi-parametric image generation framework | Scene-graph-to-image generation · COCO-Stuff, ground-truth boxes | FID: 48.6 -> 46.9 | Improved |
| Use the proposed semi-parametric image generation framework | Scene-graph-to-image generation · Visual Genome, ground-truth boxes | FID: 57.6 -> 56.7 | Improved |
| Replace PasteGAN retrieval with SCSM (ablation) | PSGIM image generation · COCO-Stuff | FID: 58.9 -> 51.6 | Improved |
| Replace PasteGAN retrieval with SCSM (ablation) | PSGIM image generation · Visual Genome | FID: 83.8 -> 63.7 | Improved |
| Replace RetrieveGAN retrieval with SCSM (ablation) | PSGIM image generation · COCO-Stuff | FID: 55.3 -> 51.6 | Improved |
| Replace RetrieveGAN retrieval with SCSM (ablation) | PSGIM image generation · Visual Genome | FID: 80.2 -> 63.7 | Improved |
| Add bounding-box position encoding to SCSM (ablation) | SCSM + PSGIM · COCO-Stuff | FID: 54.7 -> 51.6 | Improved |
| Add bounding-box position encoding to SCSM (ablation) | SCSM + PSGIM · Visual Genome | FID: 81.5 -> 63.7 | Improved |
| Use PSGIM instead of the PasteGAN generator (ablation) | SCSM-based generation · COCO-Stuff | FID: 53.5 -> 51.6 | Improved |
| Use PSGIM instead of the PasteGAN generator (ablation) | SCSM-based generation · Visual Genome | FID: 98.7 -> 63.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- DM-GAN — DM-GAN: Dynamic Memory Generative Adversarial Networks for Text-To-Image Synthesis (CVPR 2019) · better on 3 of 3 shared results
- PasteGAN — PasteGAN: A Semi-Parametric Method to Generate Image from Scene Graph (NeurIPS 2019) · better on 2 of 2 shared results
- RetrieveGAN — RetrieveGAN: Image Synthesis via Differentiable Patch Retrieval (ECCV 2020) · better on 1 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
- PasteGAN: A Semi-Parametric Method to Generate Image from Scene Graph (NeurIPS 2019 · similarity 0.88)
- Generative Photomontage (CVPR 2025 · similarity 0.84)
- ScaleCrafter: Tuning-free Higher-Resolution Visual Generation with Diffusion Models (ICLR 2024 · similarity 0.84)
- R3CD: Scene Graph to Image Generation with Relation-Aware Compositional Contrastive Control Diffusion (AAAI 2024 · similarity 0.84)
- Generating compositional scenes via Text-to-image RGBA Instance Generation (NeurIPS 2024 · similarity 0.84)
- HiCo: Hierarchical Controllable Diffusion Model for Layout-to-image Generation (NeurIPS 2024 · 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.
