arXiv cs.AI· Chuanzhi Xu, Chengkun Yue, Haolan Guo, Xuanhua Yin, Boyu Wei, Zihan Deng, Qiang Qu, Weidong Cai·· 14 小时前AI 评分23
MetaRefine：用多级特征与图像几何信息做评分精修
MetaRefine: Score Refinement with Multi-Level Features for Image Assessment
AI 导读
MetaRefine 是一种元数据感知的评分精修架构，通过轻量残差头结合图像几何信息、多级视觉特征与冻结模型的原始分数预测评分修正，不改变原有评分流程。在 21 个数据集、100 种配置上，它使 88 种配置的 PLCC 和 85 种的 SRCC 提升，平均相对增益分别为 7.55% 和 7.10%，MAE 与 RMSE 平均相对降低 23.27% 和 21.93%。
正文 · 原文
Abstract:An image's original dimensions, aspect ratio, and composition provide important cues that can affect its technical and aesthetic quality. Image assessment methods use resizing, cropping, padding, or patch sampling to accommodate different input sizes. However, these operations can alter or obscure original geometric and compositional cues. Existing assessment models also encode rich multi-level visual evidence that can be further extracted and exploited to improve existing models. This motivates us to combine visual evidence with original-image geometry for score refinement without changing the original scoring pipelines. In this paper, we introduce MetaRefine, a metadata-aware score refinement architecture whose lightweight residual heads predict score corrections from geometry, multi-level visual features, and the frozen model's original score. Across 100 configurations on 21 datasets, MetaRefine improves PLCC in 88 configurations and SRCC in 85, with average relative gains of 7.55% and 7.10%, respectively. It reduces MAE in 90 configurations and RMSE in 89, with average relative reductions of 23.27% and 21.93%, respectively. The median parameter overhead is only 1.11%. MetaRefine can further improve state-of-the-art (SOTA) assessment models on multiple benchmarks.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.40487 [cs.CV] |
| (or arXiv:2610.40487v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.40487 arXiv-issued DOI via DataCite |
Submission history
From: Chuanzhi Xu [view email]
[v1]
Thu, 29 Oct 2026 17:15:30 UTC (3,503 KB)
来源：arXiv cs.AI · arxiv.org