Research map / Image generation / Autoregressive and latent generation
A Neural Representation of Sketch Drawings
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| lower KL weight wKL from 1.00 to 0.25 (ablation) | sketch-rnn · cat | reconstruction loss LR: -0.98 -> -1.46 | Helped |
| lower KL weight wKL from 1.00 to 0.25 (ablation) | sketch-rnn · cat | KL loss LKL: 0.29 -> 1.01 | Hurt |
| lower KL weight wKL from 1.00 to 0.25 (ablation) | sketch-rnn · face | reconstruction loss LR: -1.13 -> -1.90 | Helped |
| lower KL weight wKL from 1.00 to 0.25 (ablation) | sketch-rnn · mosquito | KL loss LKL: 0.30 -> 1.54 | Hurt |
| lower KL weight wKL from 1.00 to 0.25 (ablation) | sketch-rnn · garden | reconstruction loss LR: -0.79 -> -0.99 | Helped |
| lower KL weight wKL from 1.00 to 0.50 (ablation) | sketch-rnn · firetruck | reconstruction loss LR: -1.24 -> -1.26 | No clear effect |
| add latent vector from encoder (conditional vs decoder only) (ablation) | sketch-rnn · cat | reconstruction loss LR: -0.57 -> -0.98 | Helped |
| add latent vector from encoder (conditional vs decoder only) (ablation) | sketch-rnn · mosquito | reconstruction loss LR: -0.34 -> -0.67 | Helped |
| train on multi-class (crab, face, pig, rabbit) instead of single class | sketch-rnn · crab, face, pig, rabbit | reconstruction loss LR: -0.91 vs -1.14 | Mixed |
| lower wKL from 1.00 to 0.25 in the multi-class (cat, pig) dataset (ablation) | sketch-rnn · cat, pig | KL loss LKL: 0.22 -> 0.98 | Hurt |
Closest papers by meaning
- SketchEmbedNet: Learning Novel Concepts by Imitating Drawings (ICML 2021 · similarity 0.90)
- SketchINR: A First Look into Sketches as Implicit Neural Representations (CVPR 2024 · similarity 0.90)
- SketchKnitter: Vectorized Sketch Generation with Diffusion Models (ICLR 2023 · similarity 0.90)
- SketchDNN: Joint Continuous-Discrete Diffusion for CAD Sketch Generation (ICML 2025 · similarity 0.85)
- Sketch Down the FLOPs: Towards Efficient Networks for Human Sketch (CVPR 2025 · similarity 0.85)
- It's All About Your Sketch: Democratising Sketch Control in Diffusion Models (CVPR 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.