AI-Generated Image Quality Assessment Based on Task-Specific Prompt and Multi-Granularity Similarity
These videos have been matched automatically. Contact us if they are not relevant.
These videos have been matched automatically. Contact us if they are not relevant.
View abstract on PubMed
Summary
This summary is machine-generated.A new framework, TPMS, improves AI-generated image quality assessment by using task-specific prompts and multi-level similarity. This enhances both perceptual quality and text-to-image alignment for better generative models.
Area Of Science
- Artificial Intelligence
- Computer Vision
- Machine Learning
Background
- AI-generated images (AIGIs) are gaining attention but suffer from poor quality and prompt misalignment.
- Current evaluation methods use generic prompts and coarse similarity, limiting performance.
Purpose Of The Study
- To develop an advanced framework for assessing AIGI perceptual quality and text-to-image alignment.
- To overcome limitations of existing task-agnostic evaluation methods.
Main Methods
- Proposed TPMS framework with task-specific prompts for perceptual and alignment quality.
- Implemented multi-granularity similarity: coarse-level with task-specific prompts and fine-level with initial prompts.
Main Results
- TPMS demonstrated superior performance in AIGI quality prediction.
- Extensive experiments on four benchmarks confirmed the framework's effectiveness.
Conclusions
- The proposed TPMS framework offers precise and robust AIGI quality assessment.
- Task-specific prompts and multi-granularity similarity are key to improving generative model evaluation.
These videos have been matched automatically. Contact us if they are not relevant.
These videos have been matched automatically. Contact us if they are not relevant.