Selfie Search: Image Retrieval and Face Recognition in iOS: [Invited Paper]

Abstract

This paper describes the design and development of an iOS app for selfie search, which combines face detection and recognition capabilities with content-based image retrieval techniques. The app works offline, since all processing takes place entirely on the device. It was implemented in Objective-C and it leverages functionality from Apple's Core Image API for image processing tasks and CouchbaseLite for the database layer. For face recognition, the app employs local binary patterns -- encoded as spatially enhanced histograms, with weight maps that indicate preferred areas within the cropped image containing the face. The source code is available on GitHub.