Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Compositional Zero-Shot Learning via Fine-Grained Dense Feature Composition
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 |
|---|---|---|---|
| add random feature composition (vs none) (ablation) | discriminative model trained on seen classes · DeepFashion (fine-tuned) | harmonic mean H: 5.0 -> 34.5 | Helped |
| add random feature composition (vs none) (ablation) | discriminative model · DeepFashion (fine-tuned) | zero-shot accuracy: 44.6 -> 44.9 | No clear effect |
| use prior p(H|z) for selecting related samples instead of random composition (ablation) | Composer · DeepFashion (fine-tuned) | harmonic mean H: 34.5 -> 36.5 | Helped |
| use prior p(H|z) only instead of random composition (ablation) | Composer · DeepFashion (fine-tuned) | zero-shot accuracy: 44.9 -> 43.9 | No clear effect |
| combine classification knowledge p(y|H,z) with prior p(H|z) (ablation) | Composer · DeepFashion (fine-tuned) | harmonic mean H: 36.5 -> 36.9 | Helped |
| vary sample set S during composition (vs fixed S) (ablation) | Composer · DeepFashion (fine-tuned) | harmonic mean H: 33.6 -> 36.9 | Helped |
| vary sample set S during composition (vs fixed S) (ablation) | Composer · DeepFashion (fine-tuned) | zero-shot accuracy: 46.9 -> 47.3 | Mixed |
| learned dense feature composition instead of random composition (ablation) | Random Comp · AWA2 (pre-trained) | harmonic mean H: 65.1 -> 68.8 | Helped |
| learned dense feature composition instead of random composition (ablation) | Random Comp · CUB (pre-trained) | harmonic mean H: 57.0 -> 59.9 | Helped |
| increase number of candidate combinations b (vs b=0 random composition) (ablation) | Composer · DeepFashion (fine-tuned) | zero-shot and generalized ZSL | Helped |
| increase iterations T of composition (ablation) | Composer · DeepFashion (fine-tuned) | harmonic mean H | Mixed |
Closest papers by meaning
- Beyond Seen Primitive Concepts and Attribute-Object Compositional Learning (CVPR 2024 · similarity 0.90)
- Retrieval-Augmented Primitive Representations for Compositional Zero-Shot Learning (AAAI 2024 · similarity 0.89)
- Revealing the Proximate Long-Tail Distribution in Compositional Zero-Shot Learning (AAAI 2024 · similarity 0.89)
- A Dynamic Learning Method towards Realistic Compositional Zero-Shot Learning (AAAI 2024 · similarity 0.89)
- Context-based and Diversity-driven Specificity in Compositional Zero-Shot Learning (CVPR 2024 · similarity 0.88)
- Class Semantic Attribute Perception Guided Zero-Shot Learning (AAAI 2025 · similarity 0.88)
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.