Research map / Graph learning / Knowledge and temporal graphs
Specifying Object Attributes and Relations in Interactive Scene Generation
Oron Ashual, Lior Wolf
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 perceptual loss (ablation) | scene generation model · COCO-Stuff 128x128 ablation | FID: 125.1 -> 75.4 | Improved |
| add perceptual loss (ablation) | scene generation model · COCO-Stuff 128x128 ablation | Inception score: 6.2 -> 10.4 | Improved |
| add mask discriminator loss (ablation) | scene generation model · COCO-Stuff 128x128 ablation | FID: 183.6 -> 75.4 | Improved |
| add mask discriminator loss (ablation) | scene generation model · COCO-Stuff 128x128 ablation | Inception score: 5.2 -> 10.4 | Improved |
| add image discriminator loss (ablation) | scene generation model · COCO-Stuff 128x128 ablation | FID: 122.5 -> 75.4 | Improved |
| add object discriminator loss (ablation) | scene generation model · COCO-Stuff 128x128 ablation | FID: 94.5 -> 75.4 | Improved |
| use own image discriminator instead of that of Johnson et al. (ablation) | scene generation model · COCO-Stuff 128x128 ablation | FID: 114.2 -> 75.4 | Improved |
| use own model instead of Johnson et al. (GT layout, 64x64) | scene graph to image · COCO-Stuff 64x64 | FID: 86.5 -> 48.7 | Improved |
| use own model instead of Johnson et al. (inferred layout, 64x64) | scene graph to image · COCO-Stuff 64x64 | FID: 103.4 -> 65.3 | Improved |
| vary location attributes in addition to appearance (full method) | scene generation model · COCO-Stuff diversity 64x64 | diversity score: 0.23 -> 0.43 | Improved |
| add location attributes to box prediction | layout prediction · COCO-Stuff bounding boxes | IoU: 0.41 -> 0.61 | Improved |
| use own layout prediction (without location attributes) instead of Johnson et al. | layout prediction · COCO-Stuff bounding boxes | IoU: 0.37 -> 0.41 | Improved |
Closest papers by meaning
- Scene Graph-Grounded Image Generation (AAAI 2025 · similarity 0.88)
- Scene Graph Disentanglement and Composition for Generalizable Complex Image Generation (NeurIPS 2024 · similarity 0.87)
- MMGDreamer: Mixed-Modality Graph for Geometry-Controllable 3D Indoor Scene Generation (AAAI 2025 · similarity 0.87)
- R3CD: Scene Graph to Image Generation with Relation-Aware Compositional Contrastive Control Diffusion (AAAI 2024 · similarity 0.86)
- Build-A-Scene: Interactive 3D Layout Control for Diffusion-Based Image Generation (ICLR 2025 · similarity 0.86)
- The Scene Language: Representing Scenes with Programs, Words, and Embeddings (CVPR 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.
