Research map / Continual learning / Class-incremental
Doodle It Yourself: Class Incremental Learning by Drawing a Few Sketches
Check 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 |
|---|---|---|---|
| Use sketch support with GAT-refined classifier weights (ablation) | DIY-FSCIL · Sketchy-extended, 5-shot | Acc@novel: 72.34% -> 75.84% | Helped |
| Add GAT weight refinement (ablation) | DIY-FSCIL · Sketchy-extended, 1-shot | Acc@novel: 59.93% -> 64.10% | Helped |
| Add gradient consensus during cross-domain training (ablation) | DIY-FSCIL · Sketchy-extended, stage-1 base classification | accuracy: +2.72% | Helped |
| Add stage-1 gradient consensus (ablation) | DIY-FSCIL · Sketchy-extended, 5-shot | Acc@both: 57.21% -> 60.54% | Helped |
| Add knowledge distillation loss (ablation) | DIY-FSCIL · Sketchy-extended, 1-shot | Acc@both: 51.22% -> 54.97% | Helped |
| Add knowledge distillation loss (ablation) | DIY-FSCIL · Sketchy-extended, 5-shot | Acc@both: 57.47% -> 60.54% | Helped |
| Add cross-modal episodic training with photo and sketch support (ablation) | DIY-FSCIL · Sketchy-extended, 1-shot | Acc@both: 27.67% -> 54.97% | Helped |
| Add cross-modal episodic training with photo and sketch support (ablation) | DIY-FSCIL · Sketchy-extended, 5-shot | Acc@both: 35.19% -> 60.54% | Helped |
| Use five sketch examples instead of one (ablation) | DIY-FSCIL · Sketchy-extended, 5-way | Acc@novel: 64.10% -> 75.84% | Helped |
| Increase the number of novel classes from 5-way to 10-way (ablation) | DIY-FSCIL · Sketchy-extended, 1-shot | Acc@novel: 64.10% -> 47.31% | Hurt |
| Use sketch rather than Word2Vec text support | DIY-FSCIL · Sketchy-extended, 1-shot | Acc@novel: 26.15% -> 64.10% | Helped |
| Use sketch rather than GloVe text support | DIY-FSCIL · Sketchy-extended, 1-shot | Acc@both: 22.80% -> 54.97% | Helped |
Closest papers by meaning
- Sketch Down the FLOPs: Towards Efficient Networks for Human Sketch (CVPR 2025 · similarity 0.87)
- Compositional Few-Shot Class-Incremental Learning (ICML 2024 · similarity 0.86)
- SketchFusion: Learning Universal Sketch Features through Fusing Foundation Models (CVPR 2025 · similarity 0.85)
- O3SLM: Open Weight, Open Data, and Open Vocabulary Sketch-Language Model (AAAI 2026 · similarity 0.85)
- Enhance Sketch Recognition’s Explainability via Semantic Component-Level Parsing (AAAI 2024 · similarity 0.85)
- Pre-trained Vision and Language Transformers Are Few-Shot Incremental Learners (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.