How Would Artists Design CBIR? Facilitating Collaboration to Develop Content-Based Image Retrieval on an Art-School Campus

Date of Completion

8-31-2005

Document Type

Open Access Capstone

Degree Name

Master of Arts (MA)

First Advisor

Peter Taylor

Abstract

This paper represents an art-library director’s 18-month effort to promote the development of content-based image retrieval (CBIR)–technology that searches image databases by visual elements or qualities, rather than relying only on verbal description. It traces his evolving focus: from CBIR technology, to philosophical, historical, and cognitive contexts for visual communication, to practical facilitation of collaboration among technology developers, image database developers, and visual artists. The paper offers a range of reasons why such collaboration would significantly boost CBIR development and why it could effectively occur on an art-school campus. CBIR is an electronic adaption of biological “looking,” a direct process that evolved long before verbal communication. Already supporting work in many areas, e.g., architecture, weather forecasting, and law enforcement, it could potentially benefit individuals with limited verbal communication ability, scores of professions, especially research and applied sciences, and those who create and manipulate images. Why are libraries of visual materials organized only by verbal systems, and why has CBIR research and development not involved the participation of visual artists? Since the ancient Greeks, Western society consistently has trusted rational thinking over sensory, verbal communication over visual. Examination of related issues in philosophy, cognition, and library-use behavior suggests that trust is problematic. This is exemplified by CBIR researchers’ concept of a “semantic gap” between “low-end” visual data and “high-end” verbal meaning. Product development benefits from the participation of test pilots who can push prototype capability. Artists’ skills and dispositions make them appropriate CBIR test pilots. The discipline of visual literacy, theory of multiple intelligences, and research on artists’ library searching behavior support their candidacy. Interviews with art school faculty revealed an understanding of CBIR issues and that current classroom assignments pose challenging goals for CBIR development. The author’s participation in a panel presentation at the national art libraries conference, advocating that art-school faculty and students participate in CBIR development, set the scene for a post-presentation meeting with staff of the pre-eminent ARTstor image database organization and other panel members--a CBIR developer and CBIR researcher. The meeting identified shared concerns, potential contexts for collaborating on a CBIR development project, and strategies for seeking foundation funding for that project.

Comments

Contact cct@umb.edu for access to full text

This document is currently not available here.

Share

COinS