Research map / Detection and segmentation / Open-set and few-shot detection
LayoutTransformer: Scene Layout Generation With Conceptual and Spatial Diversity
Cheng-Fu Yang, Wan-Cyuan Fan, Fu-En Yang, Yu-Chiang Frank Wang
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 LT-Net instead of CanonicalSg2Im | Scene-graph-to-layout generation · COCO-Stuff test | mIOU: 0.42 ±0.01 -> 0.49 ±0.03 | Improved |
| Use LT-Net instead of CanonicalSg2Im | Scene-graph-to-layout generation · COCO-Stuff test | Relation accuracy: 50.12 ±0.21 -> 51.36 ±0.45 | Improved |
| Use LT-Net instead of Sg2Im | Layout-to-image generation · COCO-Stuff test | FID: 69.2 ±0.4 -> 76.8 ±0.3 | Hurt |
| Use LT-Net instead of CanonicalSg2Im | Scene-graph-to-layout generation · VG-MSDN test | mIOU: 0.174 ±0.025 -> 0.183 ±0.036 | Improved |
| Use LT-Net instead of Sg2Im | Layout-to-image generation · VG-MSDN test | FID: 92.1 ±1.7 -> 99.8 ±1.7 | Hurt |
| Add object ID embedding (ablation) | LT-Net predictor · COCO-Stuff masked-word prediction | Masked accuracy: 74.45 ±0.52 -> 85.68 ±0.21 | Improved |
| Add PoP ID embedding (ablation) | LT-Net predictor · COCO-Stuff masked-word prediction | Masked accuracy: 74.45 ±0.52 -> 75.73 ±0.08 | Improved |
| Add object ID and PoP ID embeddings (ablation) | LT-Net predictor · COCO-Stuff masked-word prediction | Masked accuracy: 74.45 ±0.52 -> 87.12 ±0.17 | Improved |
| Add relation consistency loss (ablation) | LT-Net layout generator · COCO-Stuff test | Coarse-box mIOU: 0.43 ±0.01 -> 0.45 ±0.02 | Improved |
| Add relation consistency loss (ablation) | LT-Net layout generator · COCO-Stuff test | Diversity score: 0.54 ±0.03 -> 0.53 ±0.06 | No clear effect |
| Add visual-textual co-attention refinement (ablation) | LT-Net layout generator · COCO-Stuff test | FID: 85.6 ±0.3 -> 75.3 ±0.8 | Improved |
| Add GMM confidence weighting to co-attention (ablation) | LT-Net layout generator · COCO-Stuff test | FID: 75.3 ±0.8 -> 76.8 ±0.3 | Hurt |
Closest papers by meaning
- Rethinking The Training And Evaluation of Rich-Context Layout-to-Image Generation (NeurIPS 2024 · similarity 0.89)
- SSMG: Spatial-Semantic Map Guided Diffusion Model for Free-Form Layout-to-Image Generation (AAAI 2024 · similarity 0.89)
- Disentangled 3D Scene Generation with Layout Learning (ICML 2024 · similarity 0.88)
- Scene Graph Disentanglement and Composition for Generalizable Complex Image Generation (NeurIPS 2024 · similarity 0.88)
- PLACE: Adaptive Layout-Semantic Fusion for Semantic Image Synthesis (CVPR 2024 · similarity 0.87)
- SpotActor: Training-Free Layout-Controlled Consistent Image Generation (AAAI 2025 · similarity 0.87)
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.
