Research map / Vision-language and multimodal / CLIP and zero-shot
Large-Scale Few-Shot Learning via Multi-Modal Knowledge Discovery
Shuo Wang, Jun Yue, Jianzhuang Liu, Qi Tian, Meng Wang
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 semantic soft loss from textual knowledge (alpha 0.5) (ablation) | LCE classifier · NOVEL-S1 K=1 | Top-5 accuracy: 49.5 -> 51.1 | Improved |
| add semantic soft loss from textual knowledge (alpha 0.5) (ablation) | LCE classifier · NOVEL-S1 K=10 | Top-5 accuracy: 77.3 -> 78.8 | Improved |
| add semantic soft loss from textual knowledge (alpha 0.5) (ablation) | LCE classifier · ALL-S1 K=20 | Top-5 accuracy: 79.3 -> 80.3 | Improved |
| add semantic soft loss, alpha sweep 0.0 to 0.9 (ablation) | LCE classifier · NOVEL-S1 K=1 | Top-5 accuracy: 50.1 to 51.1 across alpha | Mixed |
| add imbalance control loss LIC (ablation) | LCE classifier · NOVEL-S1 K=1 | Top-5 accuracy: 49.5 -> 50.1 | Improved |
| add imbalance control loss LIC (ablation) | LCE classifier · NOVEL-S1 K=10 | Top-5 accuracy: 77.3 -> 78.1 | Improved |
| add imbalance control loss LIC (ablation) | LCE classifier · ALL-S1 K=20 | Top-5 accuracy: 79.3 -> 79.9 | Improved |
| vary imbalance control strength beta in [0,1] (ablation) | LCE + LIC · ImageNet-FS | Top-5 accuracy | No clear effect |
| concatenate foreground and background features with original features (v) (ablation) | Original-image CNN features · ImageNet-FS S1 (Fig 4) | Top-5 accuracy | Improved |
| add textual knowledge classifier to SGM features | SGM · ImNet K=1 | Top-5 accuracy: 31.4 -> 33.5 | Improved |
| add textual knowledge classifier to KTCH features | KTCH · ImNet K=1 | Top-5 accuracy: 36.0 -> 40.1 | Improved |
| add textual knowledge classifier to KTCH features | KTCH · ImNet K=5 | Top-5 accuracy: 60.4 -> 63.3 | Improved |
Closest papers by meaning
- Dynamic Few-Shot Visual Learning Without Forgetting (CVPR 2018 · similarity 0.88)
- Simple Semantic-Aided Few-Shot Learning (CVPR 2024 · similarity 0.86)
- Envisioning Class Entity Reasoning by Large Language Models for Few-shot Learning (AAAI 2025 · similarity 0.86)
- Making Large Vision Language Models to Be Good Few-Shot Learners (AAAI 2025 · similarity 0.86)
- Text Augmented Correlation Transformer For Few-shot Classification & Segmentation (CVPR 2025 · similarity 0.86)
- Few-Shot Recognition via Stage-Wise Retrieval-Augmented Finetuning (CVPR 2025 · similarity 0.85)
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.
