Research map / Label-efficient and robust learning / Recognition and long-tail
SketchAA: Abstract Representation for Abstract Sketches
Lan Yang, Kaiyue Pang, Honggang Zhang, 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 |
|---|---|---|---|
| use learnable graph aggregator instead of mean pooling (ablation) | Ours · QuickDraw recognition | top-1 accuracy: 77.41% -> 81.51% | Improved |
| use learnable graph aggregator instead of max pooling (ablation) | Ours · QuickDraw recognition | top-1 accuracy: 77.80% -> 81.51% | Improved |
| use learnable graph aggregator instead of concatenation (ablation) | Ours · QuickDraw recognition | top-1 accuracy: 78.22% -> 81.51% | Improved |
| use graph aggregator instead of relational critic (ablation) | Ours · QuickDraw recognition | top-1 accuracy: 80.42% -> 81.51% | Improved |
| use graph aggregator instead of LSTM extrapolation (ablation) | Ours · QuickDraw recognition | top-1 accuracy: 80.91% -> 81.51% | Improved |
| use graph aggregator instead of mean pooling (ablation) | Ours · QMUL Chair V2 | Acc@1: 43.24 -> 52.89 | Improved |
| use graph aggregator instead of mean pooling (ablation) | Ours · QMUL Shoe V2 | Acc@1: 29.13 -> 32.33 | Improved |
| fine-tune plain ResNet-50 instead of vector-sequence methods | ResNet-50 · QuickDraw recognition | top-1 accuracy: 78.76% vs 77.31% (SketchGCN), 78.34% (SketchFormer) | Improved |
| raise abstraction levels from q=1 to q=2 (ablation) | Ours · QuickDraw recognition | top-1 accuracy: 80.04% -> 81.51% | Improved |
| raise abstraction levels from q=2 to q=3 (ablation) | Ours · QMUL Chair V2 | Acc@1: 47.64% -> 52.89% | Improved |
| raise abstraction levels from q=3 to q=4 (ablation) | Ours · QMUL Shoe V2 | Acc@1: 32.33% -> 31.52% | Hurt |
| raise input size to 512x512 with q=4 instead of q=2 (ablation) | Ours · QuickDraw recognition | top-1 accuracy: 81.89% -> 82.05% | Improved |
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.
- SketchMate — SketchMate: Deep Hashing for Million-Scale Human Sketch Retrieval (CVPR 2018) · better on 1 of 1 shared result
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
- SEVA: Leveraging sketches to evaluate alignment between human and machine visual abstraction (NeurIPS 2023 · similarity 0.92)
- How to Handle Sketch-Abstraction in Sketch-Based Image Retrieval? (CVPR 2024 · similarity 0.91)
- SketchFusion: Learning Universal Sketch Features through Fusing Foundation Models (CVPR 2025 · similarity 0.88)
- SketchINR: A First Look into Sketches as Implicit Neural Representations (CVPR 2024 · similarity 0.87)
- Open Vocabulary Semantic Scene Sketch Understanding (CVPR 2024 · similarity 0.87)
- It's All About Your Sketch: Democratising Sketch Control in Diffusion Models (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.
