Intelligent organization of multiuser photo galleries using sub-event detection

When a group of people share photos and albums that were taken using different devices, it is ideal that these photos and albums be automatically organized to represent the different events that they belong to. The main objective is to develop a tool that uses image classification techniques to create and manage photo albums based on event recognition. It exploits Deep Convolutional Neural Networks to extract features from photos taken by multiple users. These features are then reduced and used to assess the similarities between the different photos. The similarities are used to group relevant photos using unsuper-vised clustering techniques. This provides a richer and a multi-perspective view of the events where those photos were taken. The system is tested and evaluated on a publicly available dataset. Its functionality is demonstrated in a simple web application where multiple users can use to upload photo collections and have them integrated with other users' photos to generate intelligent albums. The proposed methods proved effective as they achieved an accuracy of up to 98%.