Representative photo selection for restaurants in food blogs

Nowadays, people write comments of restaurants and upload related photos to food blogs after visiting there. Developing a mobile application which enables the user to effectively search restaurants from data in these blogs becomes an emerging need. Besides reading the comments, most people will give a glance at food photos of a restaurant and then decide whether to go or what to eat. Therefore, we propose a system to analyze and select representative photos for each restaurant based on blog-platform media. A strong food detection model is trained to retrieve food photos and an aesthetic quality assessment method is utilized to select representative photos. Based on these representative photos, users can more easily have the impression of the restaurant and review the blog in an organized way. The experimental results show that our system can generate better representative photos (i.e. much closer to the users' preferences) than existing blog platforms.