Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Relational Embedding for Few-Shot Classification
Dahyun Kang, Heeseung Kwon, Juhong Min, Minsu Cho
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 self-correlational representation (SCR) (ablation) | GAP baseline · miniImageNet | accuracy (5-way 1-shot): 65.38 -> 66.66 | Improved |
| add self-correlational representation (SCR) (ablation) | GAP baseline · CUB-200-2011 | accuracy (5-way 1-shot): 77.54 -> 78.69 | Improved |
| add cross-correlational attention (CCA) (ablation) | GAP baseline · miniImageNet | accuracy (5-way 1-shot): 65.38 -> 66.00 | Improved |
| add cross-correlational attention (CCA) (ablation) | GAP baseline · CUB-200-2011 | accuracy (5-way 1-shot): 77.54 -> 78.49 | Improved |
| add both SCR and CCA (ablation) | GAP baseline · miniImageNet | accuracy (5-way 1-shot): 65.38 -> 67.60 | Improved |
| add both SCR and CCA (ablation) | GAP baseline · CUB-200-2011 | accuracy (5-way 1-shot): 77.54 -> 79.49 | Improved |
| use SCR instead of non-local self-attention (ablation) | RENet · miniImageNet | accuracy (5-way 1-shot): 65.00 -> 66.66 | Improved |
| use channel-wise self-correlation instead of ordered cosine similarity (SCE) (ablation) | SCR · miniImageNet | accuracy (5-way 1-shot): 63.39 -> 66.66 | Improved |
| use CCA instead of CAN co-attention (ablation) | RENet · miniImageNet | accuracy (5-way 1-shot): 65.66 -> 66.00 | Improved |
| use relative instead of absolute neighborhood in self-correlation (ablation) | SCR · CUB-200-2011 | accuracy (5-way 1-shot): 76.34 -> 78.69 | Improved |
| use relative instead of absolute neighborhood in self-correlation (ablation) | SCR · miniImageNet | accuracy (5-way 1-shot): 66.41 -> 66.66 | Improved |
| use separable instead of vanilla 4D convolution in CCA (ablation) | CCA · miniImageNet / CUB | accuracy (5-way 1-shot): 65.59 -> 65.90 (mini), 78.89 -> 78.49 (CUB), GPU 60.35 -> 34.97 ms | No clear effect |
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.
- DeepEMD — DeepEMD: Few-Shot Image Classification With Differentiable Earth Mover's Distance and Structured Classifiers (CVPR 2020) · better on 5 of 6 shared results
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
- A Dual Attention Network with Semantic Embedding for Few-Shot Learning (AAAI 2019 · similarity 0.88)
- CAD: Co-Adapting Discriminative Features for Improved Few-Shot Classification (CVPR 2022 · similarity 0.87)
- Learning to Compare: Relation Network for Few-Shot Learning (CVPR 2018 · similarity 0.87)
- Rethinking Few-shot Image Classification: A Good Embedding is All You Need? (ECCV 2020 · similarity 0.87)
- A Universal Representation Transformer Layer for Few-Shot Image Classification (ICLR 2021 · similarity 0.86)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 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.
