Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Few-Shot Learning via Saliency-Guided Hallucination of Samples
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 |
|---|---|---|---|
| Add inter-class foreground-background hallucination (ablation) | SalNet · miniImageNet 5-way 1-shot | accuracy: $55.57 \pm 0.86$ -> $57.45 \pm 0.88$ | Helped |
| Add intra-class foreground-background hallucination (ablation) | SalNet · miniImageNet 5-way 5-shot | accuracy: $70.35 \pm 0.66$ -> $ 71.78\pm 0.69$ | Helped |
| Disable saliency segmentation and hallucination (ablation) | SalNet · miniImageNet 5-way 5-shot | accuracy: $ 68.87 \pm 0.67$ -> $70.35 \pm 0.66$ | Hurt |
| Add TriR regularization to intra-class hallucination (ablation) | SalNet · miniImageNet 5-way 5-shot | accuracy: $70.8\pm 0.7$ -> $71.8\pm 0.7$ | Helped |
| Add TriR regularization to inter-class hallucination (ablation) | SalNet · miniImageNet 5-way 5-shot | accuracy: $67.7\pm 0.7$ -> $67.9\pm 0.7$ | Helped |
| Use SSP instead of HSP with inter-class hallucination and TriR (ablation) | SalNet · miniImageNet 5-way 1-shot | accuracy: $56.4 \pm 0.9$ -> $57.5\pm 0.9$ | Helped |
| Dilate foreground masks to remove silhouette shapes (ablation) | SalNet Intra-class Hal. · miniImageNet 5-way 1-shot | accuracy: $55.57 \pm 0.86$ -> $\mathbf{56.67} \pm 0.85$ | Helped |
| Use inter-class hallucination rather than intra-class hallucination (ablation) | SalNet · miniImageNet 5-way 5-shot | accuracy: $ 71.78\pm 0.69$ -> $ 72.01\pm 0.67$ | Helped |
| Add inter-class feature-space hallucination | SalNet · Open MIC p1→p2, 5-way 1-shot | accuracy: 78.0 -> 79.3 | Helped |
| Add inter-class feature-space hallucination | SalNet · Open MIC p2→p3, 5-way 3-shot | accuracy: 72.6 -> 74.2 | Helped |
Closest papers by meaning
- Low-Shot Learning From Imaginary Data (CVPR 2018 · similarity 0.88)
- Adversarial Feature Hallucination Networks for Few-Shot Learning (CVPR 2020 · similarity 0.88)
- CAMERAS: Enhanced Resolution and Sanity Preserving Class Activation Mapping for Image Saliency (CVPR 2021 · similarity 0.88)
- Label Hallucination for Few-Shot Classification (AAAI 2022 · similarity 0.87)
- Learning To Memorize Feature Hallucination for One-Shot Image Generation (CVPR 2022 · similarity 0.87)
- Dense Classification and Implanting for Few-Shot Learning (CVPR 2019 · similarity 0.87)
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.