Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Transductive Few-Shot Classification on the Oblique Manifold
Guodong Qi, Huimin Yu, Zhaohui Lu, Shuzhao Li
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 transductive inference over query set (ablation) | Oblique ODC + RSSPP · mini-ImageNet 1-shot WRN28-10 | accuracy: 66.78 -> 80.64 | Improved |
| add transductive inference over query set (ablation) | Oblique ODC (no RSSPP) · tiered-ImageNet 1-shot WRN28-10 | accuracy: 71.29 -> 84.70 | Improved |
| add RSSPP (region self-attention with SPP) (ablation) | Oblique ODC, transductive · mini-ImageNet 1-shot WRN28-10 | accuracy: 80.22 -> 80.64 | Improved |
| add RSSPP (ablation) | Oblique ODC, transductive · mini-ImageNet 5-shot WRN28-10 | accuracy: 88.71 -> 89.39 | Improved |
| add RSSPP (ablation) | Oblique ODC, transductive · tiered-ImageNet 1-shot WRN28-10 | accuracy: 84.70 -> 85.22 | Improved |
| add RSSPP (ablation) | Oblique ODC, transductive · CUB 1-shot ResNet-18 | accuracy: 85.01 -> 85.87 | Improved |
| add RSSPP to Euclidean classifier with transduction (ablation) | Euclidean baseline · tiered-ImageNet 1-shot WRN28-10 | accuracy: 70.19 -> 64.26 | Hurt |
| replace Euclidean space with oblique manifold and ODC (ablation) | Euclidean baseline · tiered-ImageNet 5-shot WRN28-10 | accuracy: 85.51 -> 87.27 | Improved |
| replace Euclidean space with oblique manifold and ODC (ablation) | Euclidean baseline · mini-ImageNet 5-shot WRN28-10 | accuracy: 82.17 -> 84.34 | Improved |
| replace L2-Euclidean classifier with oblique ODC (ablation) | L2-Euclidean baseline · mini-ImageNet 1-shot WRN28-10 | accuracy: 67.61 -> 66.00 | Hurt |
| use random weight initialization instead of prototype initialization (ablation) | ODC · few-shot benchmarks | accuracy: 0.15 | Improved |
| use random anchor initialization instead of pseudo Karcher mean (ablation) | ODC · few-shot benchmarks | accuracy | Improved |
Closest papers by meaning
- Improved Few-Shot Visual Classification (CVPR 2020 · similarity 0.87)
- CAD: Co-Adapting Discriminative Features for Improved Few-Shot Classification (CVPR 2022 · similarity 0.87)
- Adaptive Poincaré Point to Set Distance for Few-Shot Classification (AAAI 2022 · similarity 0.86)
- Hubs and Hyperspheres: Reducing Hubness and Improving Transductive Few-Shot Learning With Hyperspherical Embeddings (CVPR 2023 · similarity 0.86)
- Few-Shot Classification With Feature Map Reconstruction Networks (CVPR 2021 · similarity 0.85)
- UNEM: UNrolled Generalized EM for Transductive Few-Shot Learning (CVPR 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.
