zhiyuanyou/Qwen2.5-VL-7B-GRPO-Composition-Score-Class
The zhiyuanyou/Qwen2.5-VL-7B-GRPO-Composition-Score-Class model is a 7 billion parameter vision-language model specifically fine-tuned for image composition assessment and classification. Developed by Zhiyuan You and collaborators, it excels at evaluating photographic composition quality, achieving state-of-the-art performance on datasets like CADB, GAIC, and AVA for assessment (SRCC/PLCC) and CADB for classification accuracy. This model is ideal for applications requiring automated analysis of image aesthetics and compositional correctness.
Loading preview...
PhotoFramer Composition Assessment Model
This model, developed by Zhiyuan You and collaborators, is a specialized 7 billion parameter vision-language model designed for image composition assessment and classification. It is fully fine-tuned on a comprehensive set of datasets including AVA, CADB, GAIC, and KUPCP, making it highly proficient in evaluating the aesthetic quality and compositional adherence of images.
Key Capabilities
- Composition Assessment: Evaluates image composition quality using standard metrics like SRCC (Spearman's Rank Correlation Coefficient) and PLCC (Pearson's Linear Correlation Coefficient).
- Composition Classification: Categorizes images based on their compositional attributes, providing an accuracy score.
- High Performance: Achieves leading results across multiple benchmarks, significantly outperforming other models like Qwen2.5-VL-32B, Q-Align, and AutoPhoto on datasets such as CADB, GAIC, and AVA for assessment, and CADB for classification.
Good for
- Automated photography analysis and scoring.
- Image curation and selection tools.
- Applications requiring objective evaluation of visual composition.
- Research in computational aesthetics and image quality assessment.