AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Semantic Prompt for Few-Shot Image Recognition

Wentao Chen, Chenyang Si, Zhang Zhang, Liang Wang, Zilei Wang, Tieniu Tan

CVPR 2023 · Vision-language and multimodal › Human and embodied modeling · approach: Neural image captioning · 8 citations (OpenAlex)

Read the paper

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 changedWhereMeasuredEffect
Add semantic prompts with spatial and channel interactionsVisformer-T few-shot classifier · miniImageNet 5-way 1-shotaccuracy: 65.16±0.44 -> 72.31±0.40Improved
Add semantic prompts with spatial and channel interactionsVisformer-T few-shot classifier · tieredImageNet 5-way 1-shotaccuracy: 72.38±0.50 -> 78.03±0.46Improved
Add semantic prompts with spatial and channel interactionsVisformer-T few-shot classifier · CIFAR-FS 5-way 1-shotaccuracy: 71.99±0.47 -> 82.18±0.40Improved
Add semantic prompts with spatial and channel interactionsVisformer-T few-shot classifier · FC100 5-way 1-shotaccuracy: 43.77±0.39 -> 48.53±0.38Improved
Add RandAug and RepeatAug (ablation)Visformer-T pre-training · four few-shot datasets, 1-shotaccuracy: +2.45% on averageImproved
Add spatial interaction (ablation)Semantic Prompt · four few-shot datasets, 1-shotaccuracy: 5.89% average improvementImproved
Add channel interaction (ablation)Semantic Prompt · four few-shot datasets, 1-shotaccuracy: 5.43% average improvementImproved
Combine spatial and channel interactions (ablation)Semantic Prompt · all four ablation datasets, 1-shotaccuracy: 71.59 -> 72.31; 76.20 -> 78.03; 81.19 -> 82.18; 47.83 -> 48.53Improved
Use CLIP rather than SBERT or GloVe text embeddingsSemantic Prompt · miniImageNet and tieredImageNet, 1-shotaccuracy: 70.70 -> 72.31; 73.31 -> 78.03Mixed
Use CLIP rather than SBERT or GloVe text embeddingsSemantic Prompt · miniImageNet and tieredImageNet, 5-shotaccuracy: 83.55 -> 83.42; 88.56 -> 88.55Mixed
Insert prompts in higher rather than lower feature-extractor layers (ablation)Semantic Prompt · miniImageNet and CIFAR-FS validationaccuracyMixed
Replace the linear semantic projector with an MLP (ablation)Semantic Prompt · miniImageNet 5-wayaccuracy: 72.31 -> 72.70; 83.42 -> 83.56Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy