Core Concepts
PromptIQA directly adapts to new requirements without fine-tuning, outperforming SOTA methods.
Abstract
Existing IQA methods struggle with diverse assessment requirements.
PromptIQA uses Image-Score Pairs as prompts for targeted predictions.
Trained on mixed dataset with data augmentation strategies.
Outperforms SOTA methods in performance and generalization.
Effective adaptation to new requirements with minimal data.
Stats
"PromptIQA outperforms SOTA methods with higher performance and better generalization."
"Experiments demonstrate the effectiveness of the two strategies in improving model performance and generalization."
Quotes
"PromptIQA outperforms SOTA methods with higher performance and better generalization."