Project roadmap¶
Use this page to see which models design-generators plans to port, in what order, and which ones already exist as packages. Live status such as priority, progress, and weight availability lives in each model's issue; this page changes only when a target is added or its scope changes.
Execution order¶
- Establish the shared
laygenlibrary and implement LayoutDM, because later layout packages depend on the shared output and pipeline contracts. - Implement Layout-Corrector together with LayoutGAN++, LACE, LayoutFlow, LayoutFormer++, and LayoutDiffusion, which form the next shared layout-package group.
- Implement Coarse-to-Fine, Parse-Then-Place, and Lay-Your-Scene.
- Continue with the remaining layout packages and multimodal or language-model recipe packages listed below.
- Add package-local training and conversion workflows for targets whose planned scope requires repository training.
- Add external-LLM layout-agent packages independently when their model plans are ready.
Target registry¶
The target registry below states each model's stable identity and high-level planned scope. The issue link in each row is for tracking model-specific discussion and progress; the Planned scope column is the roadmap summary.
| Target | Stable identity | Package | Planned scope | Tracking issue |
|---|---|---|---|---|
| LayoutDM | CVPR 2023, discrete layout diffusion, paper | models/layout-dm | Layout model package | issue #3 |
| LayoutGAN++ (original name: const-layout) | ACM MM 2021, deterministic layout generation, paper | models/layoutganpp | Layout model package | issue #5 |
| LACE | ICLR 2024, lightweight layout diffusion, paper | models/lace | Layout model package | issue #6 |
| LayoutFlow | ECCV 2024, flow-matching layout generation, paper | models/layout-flow | Layout model package | issue #7 |
| Layout-Corrector | ECCV 2024, LayoutDM-based correction, paper | models/layout-corrector | Layout model package | issue #8 |
| LayoutFormer++ | CVPR 2023, sequence-to-sequence layout generation, paper | models/layoutformerpp | Layout model package | issue #9 |
| Coarse-to-Fine | AAAI 2022, hierarchical layout generation, paper | models/coarse-to-fine | Layout model package | issue #10 |
| Parse-Then-Place | ICCV 2023, text-to-layout generation, paper | models/parse-then-place | Layout model package | issue #11 |
| Lay-Your-Scene | Venue: unverified, open-vocabulary scene layout generation, paper | models/layousyn | Layout model package | issue #45 |
| LT-Net (original name: layout-transformer) | CVPR 2021, scene-graph-to-layout generation, paper | models/ltnet | Layout model package | issue #12 |
| LayoutAction | AAAI 2023, autoregressive layout generation, paper | models/layout-action | Layout model package | issue #13 |
| DS-GAN (original name: PosterLayout) | CVPR 2023, content-aware poster layout generation, paper | models/ds-gan | Poster layout model package | issue #14 |
| CGB-DM (original name: Layout-DIT) | 2024 diffusion-transformer layout generation, paper | models/cgb-dm | Poster and content-aware layout model package | issue #15 |
| Flex-DM | CVPR 2023, masked multi-field layout completion, paper | models/flex-dm | Layout model package | issue #16 |
| House-GAN | ECCV 2020, graph-conditioned floorplan generation, paper | models/housegan | Floorplan layout model package | issue #17 |
| LayoutDETR | ECCV 2024, content-image layout generation, paper | models/layout-detr | Content-aware layout model package | issue #18 |
| SmartText | IEEE TMM 2021, saliency-aware text placement | models/smarttext | Content-image layout model package | issue #19 |
| LayoutDiffusion | ICCV 2023, discrete layout diffusion, paper | models/layoutdiffusion | Layout model package | issue #20 |
| RALF | CVPR 2024, retrieval-augmented layout generation, paper | models/ralf | Poster and content-aware layout model package | issue #44 |
| LayoutGAN TensorFlow | ICLR 2019, point-layout GAN, paper | — | Layout model package | issue #46 |
| PosterLLaVA | IEEE TMM 2024, multimodal poster-layout generation, paper | models/posterllava | Multimodal layout recipe package | issue #21 |
| PosterLLaMA | ECCV 2024, multimodal poster-layout generation, paper | models/posterllama | Multimodal layout recipe package | issue #22 |
| TextLap | 2024 research method, text-based graphic-layout generation, paper | — | Multimodal layout recipe package | issue #23 |
| LaDeCo | CVPR 2025, layered design generation, paper | — | Multimodal layout recipe package | issue #24 |
| LayoutNUWA | ICLR 2024, layout-as-code generation, paper | — | Multimodal layout recipe package | issue #47 |
| Uni-Layout | ACM MM 2025, unified layout generation, paper | — | Multimodal layout recipe package | issue #52 |
| Graphist | 2024 research method, graphic-design layout generation, paper | — | Multimodal layout recipe package | issue #53 |
| DLT | ICCV 2023, Diffusers-based layout generation, paper | models/dlt | Package-local training and layout conversion | issue #25 |
| Deeplayout | ICCV 2021, minGPT-based layout generation, paper | — | Package-local training and layout conversion | issue #26 |
| DogLayout | Venue: unverified, denoising diffusion GAN, paper | — | Package-local training and layout conversion | issue #27 |
| RUITE | Venue: unverified, transformer encoder layout refinement, paper | — | Package-local training and layout conversion | issue #28 |
| Desigen | CVPR 2024, background and layout generation, paper | — | Package-local training and layout conversion | issue #29 |
| LGGPT | IJCV 2025, GPT2-XL layout generation, paper | — | Package-local training and layout conversion | issue #30 |
| CanvasVAE | ICCV 2021, VAE-based layout generation, paper | models/canvas-vae | Package-local training and layout conversion | issue #31 |
| PixelVAE | ICCV 2021, visual-element encoder for CanvasVAE, paper | models/pixel-vae | Image encoder package and CPU conversion | issue #31 |
| OptiScene | NeurIPS 2025, instruction-tuned layout generation, paper | — | Package-local training and layout conversion | issue #32 |
| DirectLayout | 2025 research method, external-LLM layout generation, paper | — | External-LLM layout-agent package | issue #33 |
| SEGA | ICCV 2025, multimodal layout generation, paper | — | Package-local training and layout conversion | issue #34 |
| RADM | CIKM 2023, layout generation, paper | models/radm | Package-local training and layout conversion | issue #35 |
| GlyphDraw2 | 2024 research method, layout generation, paper | — | Package-local training and layout conversion | issue #36 |
| Layout-BLT | ECCV 2022, BLT layout generation, paper | — | Package-local training and layout conversion | issue #49 |
| Layout Generation Baselines | Baseline suite, paper | — | Package-local training and layout conversion | issue #50 |
| SciPostLayout | 2024 research method, scientific-poster layout generation, paper | — | Package-local training and layout conversion | issue #51 |
| OmniDocLayout | CVPR 2026, document layout generation, paper | — | Package-local training and layout conversion | issue #54 |
| CGL-GAN | IJCAI 2022, graphic-design layout generation, paper | — | Package-local training and layout conversion | issue #55 |
| CreatiPoster | 2025 research method, poster layout generation, paper | — | Package-local training and layout conversion | issue #56 |
| Spot-Error | Venue: unverified, non-autoregressive wireframe-conditioned layout generation, paper | — | Package-local training and layout conversion | issue #57 |
| LayoutGPT | NeurIPS 2023, external-LLM layout generation, paper | models/layout-gpt | External-LLM layout-agent package | issue #37 |
| LayoutPrompter | NeurIPS 2023, external-LLM layout generation, paper | models/layoutprompter | External-LLM layout-agent package | issue #38 |
| PosterO | CVPR 2025, language-model layout trees, paper | models/postero | External-LLM layout-agent package | issue #48 |
| GlobalMapper | ICCV 2023, 3D and urban generation, paper | — | Package-local training and layout conversion | issue #39 |
| CityGen | 2023 research method, 3D and urban generation, paper | — | Package-local training and layout conversion | issue #40 |
| COHO | ECCV 2024, context-aware 3D and urban generation, project | — | Package-local training and layout conversion | issue #41 |
| LaViGen | CVPR 2026, 3D generation, paper link unavailable | — | Package-local training and layout conversion | issue #42 |
| Imaginarium | 2025 research method, hybrid 3D generation, paper | — | Package-local training and layout conversion | issue #43 |
| LayoutVAE | ICCV 2019, label-conditioned layout generation, paper | models/layoutvae | Layout model package | — |
| Layout FID | ACM MM 2021, layout evaluation package, paper | models/layout-fid | Layout evaluation package | — |
| BASNet (SmartText support) | CVPR 2019, salient-object detection, paper | models/basnet | SmartText support package | issue #19 |
Existing owners for policy and interface decisions¶
Policy and interface decisions live with their owners: public output and conditioning contracts are in Conventions, shared package ownership and dependency direction are in Shared library architecture, and contributor workflow and lifecycle rules are in AGENTS.md. Follow those links when you need the applicable rule; the roadmap does not duplicate their contents.