Research map / Label-efficient and robust learning / Recognition and long-tail
SketchMate: Deep Hashing for Million-Scale Human Sketch Retrieval
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 |
|---|---|---|---|
| combine CNN visual features with RNN stroke-order features (ablation) | SketchMate · QuickDraw-3.8M sketch recognition | accuracy: 0.7376 -> 0.7949 | Helped |
| add the RNN stroke-order branch to the CNN branch (ablation) | SketchMate · QuickDraw-3.8M sketch recognition | accuracy: 0.7376 -> 0.7949 | Helped |
| add sketch center loss (ablation) | SketchMate with cross-entropy loss · QuickDraw-3.8M retrieval | MAP (16 bits): 0.5969 -> 0.6016 | Helped |
| add sketch center loss (ablation) | SketchMate with cross-entropy loss · QuickDraw-3.8M sketch recognition | accuracy: 0.7949 -> 0.8051 | Helped |
| replace sketch center loss with common center loss (ablation) | SketchMate with cross-entropy loss · QuickDraw-3.8M retrieval | MAP (16 bits): 0.5969 -> 0.5567 | Hurt |
| add quantization loss (ablation) | SketchMate with cross-entropy and sketch center losses · QuickDraw-3.8M retrieval | MAP (16 bits): 0.6016 -> 0.6064 | Helped |
| use the full sketch-specific hashing method instead of DLBHC | SketchMate · QuickDraw-3.8M retrieval | MAP (16 bits): 0.5453 -> 0.6064 | Helped |
| use the full sketch-specific hashing method instead of DLBHC | SketchMate · QuickDraw-3.8M retrieval | MAP (64 bits): 0.6241 -> 0.6791 | Helped |
| use the full sketch-specific hashing method instead of DLBHC | SketchMate · QuickDraw-3.8M zero-shot retrieval (20 unseen categories, 32 bits) | MAP: 0.7094 -> 0.7547 | Helped |
Closest papers by meaning
- Deep Sketch Hashing: Fast Free-Hand Sketch-Based Image Retrieval (CVPR 2017 · similarity 0.93)
- Sketch Down the FLOPs: Towards Efficient Networks for Human Sketch (CVPR 2025 · similarity 0.90)
- How to Handle Sketch-Abstraction in Sketch-Based Image Retrieval? (CVPR 2024 · similarity 0.89)
- SketchFusion: Learning Universal Sketch Features through Fusing Foundation Models (CVPR 2025 · similarity 0.88)
- ARNet: Self-Supervised FG-SBIR with Unified Sample Feature Alignment and Multi-Scale Token Recycling (AAAI 2025 · similarity 0.87)
- SketchINR: A First Look into Sketches as Implicit Neural Representations (CVPR 2024 · similarity 0.87)
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.