Research map / Graph learning / Graph self-supervised learning
SketchLattice: Latticed Representation for Sketch Manipulation
Yonggang Qi, Guoyao Su, Pinaki Nath Chowdhury, Mingkang Li, Yi-Zhe Song
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 residual connection (grid 16, nearest) (ablation) | Lattice-GCN-LSTM · QuickDraw ablation (P_mask=10%) | recognition Top-1 accuracy: 14.47% -> 40.64% | Improved |
| add residual connection (grid 8, nearest) (ablation) | Lattice-GCN-LSTM · QuickDraw ablation (P_mask=10%) | recognition Top-1 accuracy: 26.56% -> 36.16% | Improved |
| use nearby neighbors instead of nearest neighbor (grid 8) (ablation) | Lattice-GCN-LSTM · QuickDraw ablation (P_mask=10%) | recognition Top-1 accuracy: 36.16% -> 38.86% | Improved |
| use nearby neighbors instead of nearest (grid 16, P_mask=30%) (ablation) | Lattice-GCN-LSTM · QuickDraw ablation (P_mask=30%) | recognition Top-1 accuracy: 40.59% -> 39.07% | Hurt |
| increase grid n from 8 to 32 (ablation) | Lattice-GCN-LSTM · QuickDraw ablation (P_mask=10%) | recognition Top-1 accuracy: 38.86% -> 55.50% | Improved |
| increase grid n from 32 to 64 (ablation) | Lattice-GCN-LSTM · QuickDraw ablation (P_mask=10%) | recognition Top-1 accuracy: 55.50% -> 45.71% | Hurt |
| increase grid n from 32 to 64 (ablation) | Lattice-GCN-LSTM · QuickDraw ablation (P_mask=30%) | recognition Top-1 accuracy: 54.79% -> 45.07% | Hurt |
| raise corruption P_mask from 10% to 30% | Lattice-GCN-LSTM · QuickDraw sketch healing | recognition accuracy: 55.50% -> 54.79% | No clear effect |
| raise corruption P_mask from 10% to 30% | SketchRNN · QuickDraw sketch healing | recognition accuracy: 25.08% -> 3.44% | Hurt |
| remove vector stroke order input (SH to SH-VC) | SketchHealer · QuickDraw sketch healing (P_mask=10%) | sketch-to-sketch retrieval Top-1: 85.74% -> 58.48% | Hurt |
| add two distractor classes circle and clock (12 classes) | Lattice-GCN-LSTM · QuickDraw retrieval (P_mask=10%) | retrieval Top-1: 76.02% -> 77.80% | No clear effect |
| replace SketchHealer with SketchLattice on complex sketches | Lattice-GCN-LSTM · QuickDraw most complex 25 percent (P_mask=30%) | recognition accuracy: 0.37 -> 0.42 | Improved |
Closest papers by meaning
- SketchAA: Abstract Representation for Abstract Sketches (ICCV 2021 · similarity 0.88)
- SketchINR: A First Look into Sketches as Implicit Neural Representations (CVPR 2024 · similarity 0.88)
- SketchRevive: Fine-Grained Pixel-to-Vector Sketch Completion with Diffusion-Prior-Guided Multimodal LLMs (CVPR 2026 · similarity 0.88)
- Linking Sketch Patches by Learning Synonymous Proximity for Graphic Sketch Representation (AAAI 2023 · similarity 0.88)
- Sketch Down the FLOPs: Towards Efficient Networks for Human Sketch (CVPR 2025 · similarity 0.86)
- It's All About Your Sketch: Democratising Sketch Control in Diffusion Models (CVPR 2024 · 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.
