Aesthetic quality assessment in photography has gained popularity in recent years, due to the huge volume of images generated by social networks and the variety of applications this task provides. Main proposals were dependent on the data available to learn from, where datasets are composed by images with ratings from different users. Most of these works reduce the problem to a binary classification [snapshots, professional shots] using rating average.

Some articles indicate that the current forms to measure aesthetic quality present important weaknesses, since most of the ratings are concentrated in central values. This paper presents a novel solution by means of Aesthetic Ranking Value (ARV) and Weighted ARV (WARV). Both compute a score value from votes’ distribution able to determine, in an aggregated way, how (aesthetically) good an image is. In this scenario, ARV and WARV show better performance as aesthetic quality scores than the mean because they obtain more distributed and discriminatory information.

A study using state-of-the-art frameworks is conducted and a new model based on the ranking relationships is presented. All models are tested on as many tasks as aesthetic scores are available, since each score represents a new ground-truth. Although it is not directly comparable, in this study distinct measurements and statistics support the hypothesis that ARV and WARV are more homogeneously distributed between the extremes. We would like to remark that the change in the ground-truth implied by the ranking-based scores allows perfectly the use of the techniques previously designed for aesthetics quality assessment.

近年来，由于社交网络生成的大量图像以及该任务提供的各种应用，摄影中的美学质量评估已广受欢迎。主要提议依赖于可供学习的数据，其中数据集由具有不同用户评分的图像组成。这些作品中的大多数都使用平均评分将问题简化为二元分类[快照、专业镜头]。

一些文章指出，目前衡量审美质量的形式存在重大缺陷，因为大多数评级都集中在中心价值上。本文通过美学排名值（ARV）和加权ARV（WARV）提出了一种新颖的解决方案。两者都从投票分布中计算得分值，从而能够以汇总的方式确定图像的（美学）好坏程度。在这种情况下，ARV 和 WARV 在美学质量得分方面表现出比平均值更好的表现，因为它们获得了更多分布和歧视性的信息。

使用最先进的框架进行了一项研究，并提出了一种基于排名关系的新模型。所有模型都在尽可能多的任务上进行测试，因为每个分数都代表了一个新的事实。虽然不能直接比较，但在本研究中，不同的测量和统计数据支持 ARV 和 WARV 在极端之间分布更均匀的假设。我们要指出的是，基于排名的分数所暗示的基本事实的变化允许完美地使用先前为美学质量评估而设计的技术。