Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
A Closer Look at Few-shot Classification
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 |
|---|---|---|---|
| Replace the linear classifier with a cosine-distance classifier (ablation) | Baseline · CUB and mini-ImageNet, 1-shot and 5-shot, Conv-4 | accuracy | Helped |
| Use a deeper feature backbone (ablation) | Few-shot classification methods · CUB | accuracy | Mixed |
| Use a deeper feature backbone (ablation) | Baseline and meta-learning methods · mini-ImageNet 5-shot | accuracy | Mixed |
| Apply the baseline fine-tuning procedure in the cross-domain setting | Few-shot classification methods · mini-ImageNet→CUB | accuracy | Helped |
| Further adapt by training a new classifier (ablation) | MatchingNet and MAML · Few-shot scenarios, especially mini-ImageNet→CUB | accuracy | Helped |
| Further adapt by training a new classifier (ablation) | ProtoNet · Scenarios with little domain difference | accuracy | Hurt |
| Further adapt the relation module using a support/query split (ablation) | RelationNet · Few-shot scenarios | accuracy | No clear effect |
| Use 1-NN with cosine distance instead of softmax at test time (ablation) | Baseline · mini-ImageNet 1-shot | accuracy: 42.11$\pm$0.71 -> 44.18$\pm$0.69 | Helped |
| Use 1-NN with cosine distance instead of softmax at test time (ablation) | Baseline · mini-ImageNet 5-shot | accuracy: 62.53$\pm$0.69 -> 56.68$\pm$0.67 | Hurt |
| Use 1-NN with cosine distance instead of softmax at test time (ablation) | Baseline++ · mini-ImageNet 1-shot | accuracy: 48.24$\pm$0.75 -> 49.57$\pm$0.73 | Helped |
| Use 1-NN with cosine distance instead of softmax at test time (ablation) | Baseline++ · mini-ImageNet 5-shot | accuracy: 66.43$\pm$0.63 -> 61.93$\pm$0.65 | Hurt |
| Use first-order approximation instead of full MAML (ablation) | MAML · Omniglot 5-shot | validation accuracy | No clear effect |
Closest papers by meaning
- A Baseline for Few-Shot Image Classification (ICLR 2020 · similarity 0.92)
- A Closer Look at Few-shot Classification Again (ICML 2023 · similarity 0.92)
- Rethinking Generalization in Few-Shot Classification (NeurIPS 2022 · similarity 0.91)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · similarity 0.90)
- Discriminative Sample-Guided and Parameter-Efficient Feature Space Adaptation for Cross-Domain Few-Shot Learning (CVPR 2024 · similarity 0.88)
- Few-Class Arena: A Benchmark for Efficient Selection of Vision Models and Dataset Difficulty Measurement (ICLR 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.