Research map / Image generation / Fast diffusion sampling
DLT: Conditioned layout generation with Joint Discrete-Continuous Diffusion Layout Transformer
Elad Levi, Eli Brosh, Mykola Mykhailych, Meir Perez
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 |
|---|---|---|---|
| train explicitly for conditioning instead of conditioning only at inference (ablation) | DLT · PubLayNet category + size (+ half locations) | FID: 14.1 -> 2.9 | Improved |
| add condition embedding (ablation) | DLT · PubLayNet category + size (+ half locations) | FID: 4.3 -> 2.9 | Improved |
| add condition embedding when conditioning is uniform across all components (category+size) (ablation) | DLT · PubLayNet category + size | FID: 5.55 -> 5.73 | No clear effect |
| use joint continuous-discrete diffusion instead of discrete-then-continuous sequential (ablation) | DLT · PubLayNet unconditioned | FID: 50.8 -> 20.1 | Improved |
| use joint continuous-discrete diffusion instead of continuous-then-discrete sequential (ablation) | DLT · PubLayNet unconditioned | FID: 20.5 -> 20.1 | Improved |
| use DLT diffusion model with 100 steps instead of autoregressive or masked baselines | Layout generation · Magazine | inference time: ~1.65 times faster | Improved |
| use DLT diffusion model with 100 steps instead of autoregressive or masked baselines | Layout generation · RICO and PubLayNet | inference time: ~2 times slower | Hurt |
| use DLT instead of BLT on alignment score | Layout generation · PubLayNet conditioned on category | Alignment: 0.10 (BLT) vs 0.11 (DLT) | Hurt |
Closest papers by meaning
- Unifying Layout Generation With a Decoupled Diffusion Model (CVPR 2023 · similarity 0.92)
- LayoutDiffusion: Improving Graphic Layout Generation by Discrete Diffusion Probabilistic Models (ICCV 2023 · similarity 0.92)
- Towards Aligned Layout Generation via Diffusion Model with Aesthetic Constraints (ICLR 2024 · similarity 0.91)
- Visual Layout Composer: Image-Vector Dual Diffusion Model for Design Layout Generation (CVPR 2024 · similarity 0.89)
- HiCo: Hierarchical Controllable Diffusion Model for Layout-to-image Generation (NeurIPS 2024 · similarity 0.87)
- Build-A-Scene: Interactive 3D Layout Control for Diffusion-Based Image Generation (ICLR 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.
