Ranking-based scores for the assessment of aesthetic quality in photography

作者：

Highlights：

• This work presents two new ranking-based aesthetic quality scores: ARV and WARV.

• These aesthetic scores are compared to the traditional approach using the mean of votes.

• ARV and WARV prove to be more suitable as an aesthetic score by returning values that are better distributed over the entire range of values.

• The most important frameworks that solve the aesthetic quality can be easily adapted to return the proposed aesthetic scores.

摘要

•This work presents two new ranking-based aesthetic quality scores: ARV and WARV.•These aesthetic scores are compared to the traditional approach using the mean of votes.•ARV and WARV prove to be more suitable as an aesthetic score by returning values that are better distributed over the entire range of values.•The most important frameworks that solve the aesthetic quality can be easily adapted to return the proposed aesthetic scores.

论文关键词：Weakly Supervised Learning,Deep Learning,Transfer Learning,Computer vision,Ranking,Aesthetics scores

论文评审过程：Received 16 September 2021, Revised 15 June 2022, Accepted 21 June 2022, Available online 26 June 2022, Version of Record 12 July 2022.

论文官网地址：https://doi.org/10.1016/j.image.2022.116803