Portrait aesthetic scoring improves by matching photographic intent
PortraitAes: Intent-Conditioned Structured Portrait Aesthetics Assessment
Computer Vision and Pattern Recognition
Summary
The quality of a portrait photo depends on the photographer's purpose, like highlighting mood or style. Existing AI models rate photos without knowing this purpose, so they can make mistakes. The authors created a new dataset and system that considers nine different photographic goals and detailed criteria to score portraits more accurately. Their system outperforms others at judging portraits based on what the photo is trying to achieve.
What this means in practice
- •For photo platform engineers: Integrate intent-aware scoring to better filter and select user-submitted portraits according to diverse photographic goals.
- •For image generation developers: Use structured portrait quality scores to guide and improve AI-generated human images based on specific photographic intents.
Authors
Junzhou Xie, Haozhong Xiong, Xunyun Tian, Kaile Du, Tianchen Yu, Qiang Li, Wei Liu, Jiaming Liu, Ruihua Huang, Yang Shi, Guangcan Liu
Abstract
Portrait aesthetic assessment assigns comparable scores according to how effectively human-centered images fulfill their photographic intent. These scores support data filtering, candidate selection, and preference modeling in image-generation pipelines. Existing methods typically predict a single aesthetic score or use general-purpose MLLMs without conditioning on photographic intent. This omission matters because the same blur, pose, lighting, or framing choice may serve one photographic intent but undermine another. These models thus learn context-agnostic aesthetic priors and yield inconsistent, inaccurate, misleading judgments for portraits with distinct photographic objectives. We introduce PortraitAes-Bench, an 11K-scale benchmark that decomposes this task into intent-conditioned subjudgments. Expert-authored rubrics define nine photographic intents, six first-level dimensions, and 22 secondary criteria. They support a structured pipeline for intent routing, specialist assessment, verification, and score fusion. Following this structure, we train PortraitAes with multi-task supervision. We then improve score comparability through Gaussian score calibration and within-dimension cross-image ranking. On the standard benchmark, PortraitAes achieves a Pearson correlation of 0.924 and a Spearman rank correlation of 0.934. On the hard-case set, its Pearson correlation is 0.829 and its Spearman rank correlation is 0.795. Across both sets, PortraitAes outperforms the evaluated general-purpose MLLMs and specialized aesthetic baselines.