An Application for Aesthetic Quality Assessment in Photography with Interpretability Features

Area of Science:

• Computer Vision
• Image Processing

Background:

• Automatic aesthetic quality assessment is crucial for managing the vast number of images on social networks.
• Existing methods lack comprehensive analysis and user-centric features for image enhancement.

Purpose of the Study:

• To develop an application, Aesthetic Selector, for identifying high-aesthetic-quality images.
• To provide insights into the decision-making process and suggest image enhancement filters.

Main Methods:

• Analysis of existing aesthetic quality assessment proposals to identify strengths and weaknesses.
• Development of the Aesthetic Selector application incorporating an interpretability module.
• Testing the application in image selection, finding, and filter selection scenarios.

Main Results:

• Aesthetic Selector demonstrated good performance across tested scenarios.
• The interpretability module successfully identified relevant image areas influencing classification.
• The application provides a novel approach to understanding aesthetic quality decisions.

Conclusions:

• Aesthetic Selector is an innovative tool for automatic aesthetic quality assessment and image enhancement.
• The application offers unique capabilities in identifying high-quality images and explaining its reasoning.
• Interpretability enhances user trust and understanding of the automated aesthetic evaluation.