Research map / Detection and segmentation / Open-set and few-shot detection
Learning Hierarchical Semantic Image Manipulation through Structured Representations
Seunghoon Hong, Xinchen Yan, Thomas S. Huang, Honglak Lee
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 semantic layout conditioning to image-context-only generation (ablation) | SingleStream-Image · Cityscapes validation, predicted layout | Segmentation accuracy: 59.6 -> 76.3 | Improved |
| Add image-context conditioning to layout-only generation (ablation) | SingleStream-Layout · Cityscapes validation, predicted layout | SSIM: 0.268 -> 0.296 | Improved |
| Use separate image and layout encoding pathways instead of one concatenated pathway (ablation) | Image generator · Cityscapes validation, predicted layout | Human eval. preference: 29.9 -> 40.0 | Improved |
| Use separate image and layout encoding pathways instead of one concatenated pathway (ablation) | Image generator · Cityscapes validation, predicted layout | SSIM: 0.296 -> 0.299 | Improved |
| Use predicted rather than ground-truth semantic layout (ablation) | TwoStream image generator · Cityscapes validation | Perceptual quality | Hurt |
| Use predicted layout conditioning rather than image context alone (ablation) | TwoStream-Pred · Cityscapes validation | Image generation quality | Improved |
| Use hierarchical manipulation instead of copying and pasting training objects | Image manipulation · Cityscapes validation, AMT comparison | Human preference: 76.33% | Improved |
| Add segmentation layout as input to Context Encoder | Context Encoder · Cityscapes validation | Segmentation accuracy: 26.4 -> 35.7 | Improved |
Closest papers by meaning
- Semantic Image Manipulation Using Scene Graphs (CVPR 2020 · similarity 0.89)
- Semantic Layout Manipulation With High-Resolution Sparse Attention (TPAMI 2022 · similarity 0.88)
- Structure-Guided Image Completion With Image-Level and Object-Level Semantic Discriminators (TPAMI 2024 · similarity 0.87)
- Learning to Manipulate Artistic Images (AAAI 2024 · similarity 0.86)
- Rethinking The Training And Evaluation of Rich-Context Layout-to-Image Generation (NeurIPS 2024 · similarity 0.86)
- HiCo: Hierarchical Controllable Diffusion Model for Layout-to-image Generation (NeurIPS 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.
