Personal photo organization

Abstract

We live in a society were a large portion of the population has, or has used a smartphone or/and a digital camera. In addition, hard drives and cloud storage are getting cheaper and cheaper, leading to a tremendous growth in stored personal photos. Organizing manually such large collections of photographs would be unfeasible. To solve the need of organizing such a large collection of personal photos automatically, we propose a system that first estimates the user profile, and then classifies the photos in categories related to the estimated profile. To estimate the user profile, we adapt a topic discovery method used in document analysis, called Probabilistic Latent Semantic Analysis (pLSA). Based on the estimated profile, we provide a personalized photo classification, using a set of pretrained Convolutional Neuronal Network (CNN). In this manner, we achieve an efficient, scalable, flexible and more tailored organization of the user’s photos.