Identifying Professional Photographers Through Image Quality and Aesthetics in Flickr

TL;DR

This study introduces a large labeled Flickr dataset to classify professional photographers using multimodal features, analyzing photo quality, aesthetics, and social activity to distinguish expertise levels.

Contribution

It creates and open-sources one of the largest Flickr datasets with multimodal data and demonstrates the feasibility of predicting professional status based on various features.

Findings

Feasibility of predicting professional photographers using machine learning.

Correlation between photo aesthetics, technical quality, and social activity.

Characteristics that differentiate professionals from amateurs.

Abstract

In our generation, there is an undoubted rise in the use of social media and specifically photo and video sharing platforms. These sites have proved their ability to yield rich data sets through the users' interaction which can be used to perform a data-driven evaluation of capabilities. Nevertheless, this study reveals the lack of suitable data sets in photo and video sharing platforms and evaluation processes across them. In this way, our first contribution is the creation of one of the largest labelled data sets in Flickr with the multimodal data which has been open sourced as part of this contribution. Predicated on these data, we explored machine learning models and concluded that it is feasible to properly predict whether a user is a professional photographer or not based on self-reported occupation labels and several feature representations out of the user, photo and crowdsourced sets. We also examined the relationship between the aesthetics and technical quality of a picture and the social activity of that picture. Finally, we depicted which characteristics differentiate professional photographers from non-professionals. As far as we know, the results presented in this work represent an important novelty for the users' expertise identification which researchers from various domains can use for different applications.