Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
One-Shot Image Classification by Learning to Restore Prototypes
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 |
|---|---|---|---|
| Restore noisy support prototypes with RestoreNet (ablation) | DEML+Prototypical Nets · miniImageNet and seven N-way settings | 1-shot accuracy: 1.89; 1.44; 1.16; 1.06; 1.05; 1.04; 0.79 | Helped |
| Apply self-training to the baseline (ablation) | DEML+Prototypical Nets · miniImageNet, 5-way | 1-shot accuracy: 57.63±0.20 -> 59.78±0.22 | Helped |
| Add prototype restoration to self-training (ablation) | Self-training + Baseline · miniImageNet, 5-way | 1-shot accuracy: 59.78±0.22 -> 61.14±0.22 | Helped |
| Use only λ=100 farthest training images per class rather than λ=200–600 (ablation) | RestoreNet · miniImageNet, 5-way | 1-shot accuracy: 59.28±0.20; 58.93±0.20; 59.12±0.20; 58.84±0.20; 58.55±0.20; 58.21±0.20 | Helped |
| Use a one-unit hidden layer in the regressor (ablation) | RestoreNet · miniImageNet, 5-way | 1-shot accuracy: 57.67±0.83 -> 58.98±0.83 | Helped |
| Restore prototypes on CIFAR-100 | DEML+Prototypical Nets · CIFAR-100 | 5-way 1-shot accuracy | Helped |
| Add self-training to RestoreNet | RestoreNet · CIFAR-100, Caltech-256, and CUB-200 | 5-way 1-shot accuracy | Helped |
Closest papers by meaning
- Matching Networks for One Shot Learning (NeurIPS 2016 · similarity 0.88)
- CLEAR: Cumulative LEARning for One-Shot One-Class Image Recognition (CVPR 2018 · similarity 0.88)
- One-Shot Metric Learning for Person Re-Identification (CVPR 2017 · similarity 0.87)
- TADAM: Task dependent adaptive metric for improved few-shot learning (NeurIPS 2018 · similarity 0.86)
- A Baseline for Few-Shot Image Classification (ICLR 2020 · similarity 0.86)
- Restoring Negative Information in Few-Shot Object Detection (NeurIPS 2020 · similarity 0.86)
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.