Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Tree Structure-Aware Few-Shot Image Classification via Hierarchical Aggregation
Min Zhang, Siteng Huang, Wenbin Li, Donglin Wang
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 HTS (tree-structured pretext aggregation, SSL variant) | ProtoNet Conv4-64 · miniImageNet 5-way 1-shot | accuracy (%): 44.42 -> 58.96 | Improved |
| add HTS (SSL variant) | ProtoNet Conv4-64 · miniImageNet 5-way 5-shot | accuracy (%): 64.24 -> 75.17 | Improved |
| add HTS (SSL variant) | ProtoNet Conv4-64 · CUB-200-2011 5-way 1-shot | accuracy (%): 51.31 -> 67.32 | Improved |
| add HTS (SSL variant) | GNN Conv4-64 · tieredImageNet 5-way 5-shot | accuracy (%): 55.31 -> 70.42 | Improved |
| add HTS (SSL variant) | MatchingNet Conv4-64 · miniImageNet 5-way 1-shot | accuracy (%): 48.14 -> 52.29 | Mixed |
| use separate label spaces (HTS_SSL) instead of single label space (HTS_DA) (ablation) | ProtoNet+HTS Conv4-64 · miniImageNet 5-way 1-shot | accuracy (%): 57.39 -> 58.96 | Improved |
| use separate label spaces (HTS_SSL) instead of single label space (HTS_DA) (ablation) | MatchingNet+HTS Conv4-64 · miniImageNet 5-way 1-shot | accuracy (%): 53.64 -> 52.29 | Hurt |
| add HTS across domains | ProtoNet Conv4-64 · tieredImageNet->miniImageNet 5-way 1-shot | accuracy (%): 47.01 -> 55.29 | Improved |
| add HTS across domains | ProtoNet Conv4-64 · miniImageNet->CUB-200-2011 5-way 1-shot | accuracy (%): 32.18 -> 39.57 | Improved |
| add HTS across domains | ProtoNet Conv4-64 · tieredImageNet->CUB-200-2011 5-way 5-shot | accuracy (%): 56.58 -> 60.71 | Improved |
| add plain pretext-task data augmentation (single label space, ERM) without HTS | ProtoNet · miniImageNet, CUB, CIFAR-FS, tieredImageNet 5-shot | accuracy (%) | Hurt |
| add plain pretext-task self-supervised auxiliary losses without HTS | ProtoNet · miniImageNet, CUB, CIFAR-FS, tieredImageNet 5-shot | accuracy (%) | Mixed |
Closest papers by meaning
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · similarity 0.89)
- TAFSSL: Task-Adaptive Feature Sub-Space Learning for few-shot classification (ECCV 2020 · similarity 0.89)
- Rethinking Few-shot Image Classification: A Good Embedding is All You Need? (ECCV 2020 · similarity 0.88)
- A Closer Look at Few-shot Classification Again (ICML 2023 · similarity 0.88)
- LibFewShot: A Comprehensive Library for Few-Shot Learning (TPAMI 2023 · similarity 0.88)
- Dynamic Few-Shot Visual Learning Without Forgetting (CVPR 2018 · 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.
