Research map / Image generation / Autoregressive and latent generation
ShapeAR: Generating Editable Shape Layers via Autoregressive Diffusion
Souymodip Chakraborty, Ankur Singh, Amit Vikram Singh, Vineet Batra, Ankit Phogat
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 |
|---|---|---|---|
| increase layer count from 4 to 12 (ablation) | ShapeAR · Held-out MMSVG/OmniSVG | 1-SSIM: 0.027 -> 0.0144 | Improved |
| increase layer count from 4 to 12 (ablation) | ShapeAR · Held-out MMSVG/OmniSVG | L1: 3.1e−5 -> 2.2e−5 | Improved |
| increase layer count from 4 to 12 (ablation) | ShapeAR · Held-out MMSVG/OmniSVG | Blended shapes: 0.0511 -> 0.0013 | Improved |
| increase layer count from 8 to 12 (ablation) | ShapeAR · Held-out MMSVG/OmniSVG | Redundant shape count: 0.0077 -> 0.0078 | Hurt |
| vary layer count (4, 8, 12) (ablation) | ShapeAR · Held-out MMSVG/OmniSVG | convexity and symmetry | No clear effect |
| use ShapeAR instead of Adobe Illustrator vectorization | vectorization · Held-out MMSVG/OmniSVG | Convexity score: 0.68 -> 0.88 | Improved |
| use ShapeAR instead of Adobe Illustrator vectorization | vectorization · Held-out MMSVG/OmniSVG | Symmetry score: 0.29 -> 0.49 | Improved |
| use ShapeAR instead of Adobe Illustrator vectorization | vectorization · Held-out MMSVG/OmniSVG | KID: 0.388 -> 0.374 | Improved |
| use ShapeAR instead of LIVSS | vectorization · Vectorization benchmark (100 images) | 1-SSIM: 0.057 -> 0.020 | Improved |
| use ShapeAR instead of LIVSS | vectorization · Vectorization benchmark (100 images) | LPIPS: 0.049 -> 0.014 | Improved |
| use ShapeAR instead of LIVSS | vectorization · Vectorization benchmark (100 images) | L1: 0.008 -> 0.005 | Improved |
| increase image complexity (number of paths) | ShapeAR and Illustrator baseline · Held-out MMSVG/OmniSVG | L1 and SSIM reconstruction error | Hurt |
Closest papers by meaning
- PixelART: Image-to-Layer Decomposition without Latents or Text-to-Image Pretraining (NeurIPS 2026 · similarity 0.86)
- Sin3DM: Learning a Diffusion Model from a Single 3D Textured Shape (ICLR 2024 · similarity 0.86)
- Single Mesh Diffusion Models with Field Latents for Texture Generation (CVPR 2024 · similarity 0.86)
- EditAR: Unified Conditional Generation with Autoregressive Models (CVPR 2025 · similarity 0.86)
- GPLD3D: Latent Diffusion of 3D Shape Generative Models by Enforcing Geometric and Physical Priors (CVPR 2024 · similarity 0.86)
- Vector Grimoire: Codebook-based Shape Generation under Raster Image Supervision (ICML 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.
