Research map / Image generation / Autoregressive and latent generation
Image Synthesis From Layout With Locality-Aware Mask Adaption
Zejian Li, Jingyu Wu, Immanuel Koh, Yongchuan Tang, Lingyun Sun
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 Locality-Aware Mask Adaption (ablation) | LAMA model · COCO-Stuff 128x128 | FID: (+3.72) 27.57 vs 23.85 | Improved |
| add Locality-Aware Mask Adaption (ablation) | LAMA model · COCO-Stuff 128x128 | SceneFID: (+4.63) 16.92 vs 12.35 | Improved |
| add LAMA module (ablation) | LostGAN-V1 · COCO-Stuff 128x128 | FID: 29.65 -> 26.13 | Improved |
| add LAMA module (ablation) | LostGAN-V1 · COCO-Stuff 128x128 | SceneFID: 20.03 -> 22.74 | Hurt |
| use local neighborhood query in LAMA instead of pixel query only (ablation) | LAMA model · COCO-Stuff 128x128 | FID: (+5.67) 29.52 vs 23.85 | Improved |
| use Batch-Group Normalization instead of BN only (ablation) | LAMA model · COCO-Stuff 128x128 | FID: (+0.64) 24.49 vs 23.85 | Improved |
| use Batch-Group Normalization instead of GN only (ablation) | LAMA model · COCO-Stuff 128x128 | FID: (+1.32) 25.17 vs 23.85 | Improved |
| add noise injection (ablation) | LAMA model · COCO-Stuff 128x128 | FID: (+1.33) 25.18 vs 23.85 | Improved |
| add extra reconstruction loss (ablation) | LAMA model · COCO-Stuff 128x128 | FID: 23.85 -> 30.97 | Hurt |
| train mask-to-image with ground-truth masks (ablation) | LAMA model · COCO-Stuff 128x128 | AP: (+5.7%) 13.6% | Improved |
| train mask-to-image with ground-truth masks (ablation) | LostGAN-V1 · COCO-Stuff 128x128 | FID: (-8.38) 21.27 | Improved |
| LAMA vs previous state of the art | LAMA model · COCO-Stuff 256x256 | FID: 41.65 -> 31.12 | Improved |
Closest papers by meaning
- OverLayBench: A Benchmark for Layout-to-Image Generation with Dense Overlaps (NeurIPS 2025 · similarity 0.88)
- PLACE: Adaptive Layout-Semantic Fusion for Semantic Image Synthesis (CVPR 2024 · similarity 0.86)
- SmartMask: Context Aware High-Fidelity Mask Generation for Fine-grained Object Insertion and Layout Control (CVPR 2024 · similarity 0.86)
- Click2Mask: Local Editing with Dynamic Mask Generation (AAAI 2025 · similarity 0.85)
- OverLay++: Dense-Overlap Layout-to-Image Generation Dataset (NeurIPS 2026 · similarity 0.85)
- InstanceAssemble: Layout-Aware Image Generation via Instance Assembling Attention (NeurIPS 2025 · 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.
