Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Few-Shot Open-Set Recognition by Transformation Consistency
Check 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 |
|---|---|---|---|
| Use transformation-consistency detection instead of direct prototype distance (ablation) | FEAT-based SnaTCHer · tieredImageNet 5-way 1-shot | AUROC: 69.36 -> 74.28 | Helped |
| Use transformation-consistency detection instead of direct prototype distance (ablation) | FEAT-based SnaTCHer · tieredImageNet 5-way 5-shot | AUROC: 77.63 -> 82.02 | Helped |
| Use transformation-consistency detection instead of classification probability (ablation) | FEAT-based SnaTCHer · tieredImageNet 5-way 1-shot | AUROC: 63.60 -> 74.28 | Helped |
| Use transformation-consistency detection instead of direct prototype distance (ablation) | LN-based SnaTCHer · tieredImageNet 5-way 1-shot | AUROC: 68.92 -> 74.07 | Helped |
| Use transformation-consistency detection instead of direct prototype distance (ablation) | IN-based SnaTCHer · tieredImageNet 5-way 1-shot | AUROC: 61.89 -> 71.22 | Helped |
| Use transformation-consistency detection instead of direct prototype distance (ablation) | TaskNorm-based SnaTCHer · tieredImageNet 5-way 1-shot | AUROC: 69.72 -> 74.84 | Helped |
| Use transformation-consistency detection instead of direct prototype distance (ablation) | LTN-based SnaTCHer · tieredImageNet 5-way 1-shot | AUROC: 69.80 -> 74.95 | Helped |
| Add transformation-consistency detection to identity-transformed prototypes (ablation) | Identity-based SnaTCHer · tieredImageNet 5-way 1-shot | AUROC: 72.40 -> 72.40 | No clear effect |
| Replace FEAT prototype transformation with DeepSets (ablation) | SnaTCHer · tieredImageNet 5-way 1-shot | AUROC: 74.28 -> 68.21 | Hurt |
| Replace FEAT prototype transformation with LN (ablation) | SnaTCHer · tieredImageNet 5-way 1-shot | AUROC: 74.28 -> 74.07 | Mixed |
| Fuse instance and layer normalization with task-adaptive weights (ablation) | SnaTCHer-L · tieredImageNet 5-way 1-shot | AUROC: 74.07 -> 74.95 | Helped |
| Use SnaTCHer-L instead of FEAT | miniImageNet 5-way 1-shot · miniImageNet 5-way 1-shot | AUROC: 57.01±0.84 -> 69.40±0.92 | Helped |
Closest papers by meaning
- Meta Evidential Transformer for Few-Shot Open-Set Recognition (ICML 2024 · similarity 0.91)
- The Devil is in the Wrongly-classified Samples: Towards Unified Open-set Recognition (ICLR 2023 · similarity 0.90)
- Few-Shot Open-Set Recognition Using Meta-Learning (CVPR 2020 · similarity 0.89)
- All Beings Are Equal in Open Set Recognition (AAAI 2024 · similarity 0.86)
- Frequency Shuffling and Enhancement for Open Set Recognition (AAAI 2024 · similarity 0.85)
- FSL-Rectifier: Rectify Outliers in Few-Shot Learning via Test-Time Augmentation (AAAI 2025 · similarity 0.84)
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.