Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Binocular Mutual Learning for Improving Few-Shot Classification
Ziqi Zhou, Xi Qiu, Jiangtao Xie, Jianan Wu, Chi Zhang
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 mutual KL loss between global and local views (ablation) | BML · miniImageNet 5-shot | accuracy: 82.84 -> 83.63 | Improved |
| add mutual KL loss to local branch (ablation) | BML-Local · miniImageNet 5-shot | accuracy: 79.29 -> 80.95 | Improved |
| add elastic loss to local branch (ablation) | Baseline-local · miniImageNet 5-shot | accuracy: 77.07 -> 77.77 | Improved |
| share first three blocks instead of none (ablation) | BML · miniImageNet 5-shot | accuracy: 83.10 -> 83.63 | Improved |
| train global and local views jointly with mutual learning instead of simple ensemble (ablation) | BML · miniImageNet 5-shot | accuracy: 81.08 -> 83.10 | Improved |
| add binocular training to local branch (ablation) | Baseline-local · miniImageNet | accuracy: 1.4%-9.7% | Improved |
| add binocular training to global branch (ablation) | Baseline-global · miniImageNet | accuracy: 1.8%-13.3% | Improved |
| add binocular learning over single global view (ablation) | Baseline-global · FC100 5-shot | accuracy: 61.03 -> 63.03 | Improved |
| combine global and local views (ablation) | Baseline-global · CUB-200-2011 5-way 1-shot | accuracy: 60.13 -> 76.21 | Improved |
| enlarge test image size to 224x224 (robustness) (ablation) | BML vs DeepEMD · miniImageNet | accuracy: 83.59 -> 81.57 | No clear effect |
| apply Gaussian blur to test images (ablation) | BML vs Rethink-Distill · miniImageNet | accuracy: 83.59 -> 61.96 | Mixed |
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 4 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
- CAD: Co-Adapting Discriminative Features for Improved Few-Shot Classification (CVPR 2022 · similarity 0.87)
- Learning To Affiliate: Mutual Centralized Learning for Few-Shot Classification (CVPR 2022 · similarity 0.87)
- Making Large Vision Language Models to Be Good Few-Shot Learners (AAAI 2025 · similarity 0.86)
- Few-Class Arena: A Benchmark for Efficient Selection of Vision Models and Dataset Difficulty Measurement (ICLR 2025 · similarity 0.85)
- Rethinking Few-Shot Adaptation of Vision-Language Models in Two Stages (CVPR 2025 · similarity 0.85)
- What Makes Good Collaborative Views? Contrastive Mutual Information Maximization for Multi-Agent Perception (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.
