Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
FD-Align: Feature Discrimination Alignment for Fine-tuning Pre-Trained Models in Few-Shot Learning
Kun Song, Huimin Ma, Bochao Zou, Huishuai Zhang, Weiran Huang
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 FD-Align fine-tuning before APE-T | CLIP ViT-B/32 · ImageNetV2 | accuracy: 58.31 -> 60.83 | Improved |
| Add FD-Align fine-tuning before APE | CLIP ViT-B/32 · ImageNet-Sketch | accuracy: 43.28 -> 44.23 | Improved |
| Add FD-Align fine-tuning to the visual encoder | CLIP, 5-way-1-shot · Sketch | accuracy: 73.84±0.56 -> 79.27±0.38 | Improved |
| Add FD-Align fine-tuning to the visual encoder | CLIP, 5-way-5-shot · Traffic Signs | accuracy: 78.11±0.24 -> 73.39±0.29 | Hurt |
| Add FD-Align fine-tuning to the visual encoder | CLIP, 5-way-1-shot · EuroSAT | accuracy: 63.99±0.39 -> 60.39±0.43 | Hurt |
| Add FD-Align fine-tuning to the visual encoder | CLIP, 5-way-1-shot · Mini-test | accuracy: 93.55±0.17 -> 95.04±0.18 | Improved |
| Replace the full 80-template prototype set with SPC-filtered and clustered prototypes (ablation) | FD-Align · 11-dataset average, 16-shot | accuracy: 74.03 -> 74.38 | Improved |
| Replace Tip-Adapter templates with SPC-filtered and clustered prototypes (ablation) | FD-Align · 11-dataset average, 16-shot | accuracy: 66.30 -> 74.38 | Improved |
| Replace Tip-Adapter templates with SPC-filtered and clustered prototypes (ablation) | FD-Align · 11-dataset average, 1-shot | accuracy: 61.14 -> 63.92 | Improved |
Closest papers by meaning
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · similarity 0.89)
- Strong Baselines for Parameter-Efficient Few-Shot Fine-Tuning (AAAI 2024 · similarity 0.88)
- Learning Dynamic Alignment via Meta-Filter for Few-Shot Learning (CVPR 2021 · similarity 0.87)
- Neural Fine-Tuning Search for Few-Shot Learning (ICLR 2024 · similarity 0.86)
- Cross-Layer and Cross-Sample Feature Optimization Network for Few-Shot Fine-Grained Image Classification (AAAI 2024 · similarity 0.86)
- Twofold Debiasing Enhances Fine-Grained Learning with Coarse Labels (AAAI 2025 · 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.
