Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Matching Feature Sets for Few-Shot Image Classification
Arman Afrasiyabi, Hugo Larochelle, Jean-François Lalonde, Christian Gagné
Check 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 |
|---|---|---|---|
| Use Sum-min set matching instead of Match-sum (ablation) | SetFeat · miniImageNet 5-way 1-shot | accuracy: 56.50 -> 59.10 | Improved |
| Use Sum-min set matching instead of Match-sum (ablation) | SetFeat · tieredImageNet 5-way 1-shot | accuracy: 71.22 -> 73.63 | Improved |
| Use Sum-min set matching instead of Match-sum (ablation) | SetFeat4-64 · CUB 5-way 5-shot | accuracy: 83.82 -> 87.05 | Improved |
| Place mappers across multiple backbone stages instead of all after the final block (ablation) | SetFeat4-64 · miniImageNet validation 5-way 1-shot | accuracy: 52.90 -> 54.71 | Improved |
| Preserve separate feature sets rather than concatenating mapper outputs into one vector (ablation) | SetFeat4-64 · miniImageNet validation | accuracy: 54.71 -> 53.56 | Improved |
| Increase the number of summed mapper distances from top-1 to top-10 (ablation) | SetFeat4 on CUB · CUB validation 5-way 5-shot | accuracy: 84.94 -> 87.05 | Improved |
| Replace Sum-min with Hausdorff distance (ablation) | SetFeat4-64 · miniImageNet 5-way 1-shot | accuracy: 57.18 -> 56.07 | Hurt |
| Replace Sum-min with Hausdorff distance (ablation) | SetFeat4-64 · CUB 5-way 5-shot | accuracy: 87.05 -> 84.85 | Hurt |
| Add FC layers to match SetFeat4-64's parameter count (ablation) | Conv4-64 · miniImageNet 5-way 1-shot | accuracy: 49.42 -> 49.98 | Improved |
| Use SetFeat's set representation rather than a parameter-matched augmented Conv4-64 (ablation) | few-shot classifier · miniImageNet 5-way 1-shot | accuracy: 49.98 -> 57.18 | Improved |
| Use SetFeat's set representation rather than a parameter-matched augmented Conv4-64 (ablation) | few-shot classifier · CUB 5-way 1-shot | accuracy: 69.11 -> 73.50 | Improved |
Closest papers by meaning
- Few-Shot Learning via Embedding Adaptation With Set-to-Set Functions (CVPR 2020 · similarity 0.89)
- Dense Classification and Implanting for Few-Shot Learning (CVPR 2019 · similarity 0.88)
- Revisiting Local Descriptor Based Image-To-Class Measure for Few-Shot Learning (CVPR 2019 · similarity 0.88)
- Impact of base dataset design on few-shot image classification (ECCV 2020 · similarity 0.88)
- A Baseline for Few-Shot Image Classification (ICLR 2020 · similarity 0.88)
- CAD: Co-Adapting Discriminative Features for Improved Few-Shot Classification (CVPR 2022 · 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.
