Research map / Image generation / Autoregressive and latent generation
VQ-SGen: A Vector Quantized Stroke Representation for Creative Sketch Generation
Jiawei Wang, Zhiming Cui, Changjian Li
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 vector quantization of stroke representation (ablation) | VQ-SGen · Creative Birds | FID: 48.53 -> 15.78 | Improved |
| add vector quantization of stroke representation (ablation) | VQ-SGen · Creative Creatures | FID: 54.56 -> 17.61 | Improved |
| decouple stroke shape and position codebooks (ablation) | VQ-SGen · Creative Birds | FID: 17.14 -> 15.78 | Improved |
| decouple stroke shape and position codebooks (ablation) | VQ-SGen · Creative Creatures | FID: 19.42 -> 17.61 | Improved |
| add label Transformer (ablation) | VQ-SGen · Creative Birds | FID: 16.51 -> 15.78 | Improved |
| add label Transformer (ablation) | VQ-SGen · Creative Creatures | FID: 19.21 -> 17.61 | Improved |
| increase codebook size from 2048 to 8192 (dim 512) (ablation) | VQ-SGen · Creative Creatures | FID: 43.21 -> 17.61 | Improved |
| increase codebook size from 4096 to 8192 (dim 512) (ablation) | VQ-SGen · Creative Birds | FID: 16.92 -> 15.78 | Improved |
| increase feature dimension from 512 to 1024 at size 4096 (ablation) | VQ-SGen · Creative Birds | FID: 16.92 -> 16.67 | Improved |
| use bitmap CNN stroke encoder instead of Sketchformer point sequence encoder (ablation) | VQ-SGen · Creative Birds | FID: 75.42 -> 15.78 | Improved |
| use VQ-SGen instead of DoodleFormer | VQ-SGen · Creative Creatures | FID: 20.43 -> 17.61 | Improved |
| use VQ-SGen instead of DoodlerGAN on CS for Creative Birds | VQ-SGen · Creative Birds | CS: 0.69 vs 0.53 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DoodleFormer — DoodleFormer: Creative Sketch Drawing with Transformers (ECCV 2022) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- StrokeFusion: Vector Sketch Generation via Joint Stroke-UDF Encoding and Latent Sequence Diffusion (AAAI 2026 · similarity 0.91)
- BézierSketch: A generative model for scalable vector sketches (ECCV 2020 · similarity 0.90)
- AI-Sketcher : A Deep Generative Model for Producing High-Quality Sketches (AAAI 2019 · similarity 0.89)
- SketchINR: A First Look into Sketches as Implicit Neural Representations (CVPR 2024 · similarity 0.87)
- Vector Grimoire: Codebook-based Shape Generation under Raster Image Supervision (ICML 2025 · similarity 0.86)
- StrokeNUWA—Tokenizing Strokes for Vector Graphic Synthesis (ICML 2024 · similarity 0.85)
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.
