SHOEBOX: A DIGITAL PHOTO MANAGEMENT SYSTEM
Timothy J. Mills, David Pye, D. Sinclair · 2000
This paper reports recent work at AT&T to develop a system for the management of personal digital photograph collections. Shoebox, the resulting software package, provides a range of browsing and searching facilities, utilising spoken annotations and image content to enable both semantically similar and visually similar images to be retrieved. We report on the design of the system, the construction of a test collection, and the evaluation of its searching facilities. The results show that audio annotation is an effective means of retrieval for photographs, which significantly out-performs image content-based techniques. Keywords: speech indexing, image retrieval, test collection development for MMIR. 1. INTRODUCTION The DART (Digital Asset Retrieval Technology) project [1] at AT&T Laboratories Cambridge is concerned with management of digital media such as text and hypertext documents, images, audio and video recordings. DART aims to provide the means to index, annotate, navigate an...