"This is clearly the most comprehensive and thoughtful compendium of knowledge on language/vision integration out there, and I'm sure it will be a valuable resource to many researchers and instructors."
- Sven Dickinson, Series Editor (University of Toronto)
Modeling data from visual and linguistic modalities together creates opportunities for better understanding of both, and supports many useful applications. Examples of dual visual-linguistic data includes images with keywords, video with narrative, and figures in documents. We consider two key task-driven themes: translating from one modality to another (e.g., inferring annotations for images) and understanding the data using all modalities, where one modality can help disambiguate information in another. The multiple modalities can either be essentially semantically redundant (e.g., keywords provided by a person looking at the image), or largely complementary (e.g., meta data such as the camera used). Redundancy and complementarity are two endpoints of a scale, and we observe that good performance on translation requires some redundancy, and that joint inference is most useful where some information is complementary.
Computational methods discussed are broadly organized into ones for simple keywords, ones going beyond keywords toward natural language, and ones considering sequential aspects of natural language. Methods for keywords are further organized based on localization of semantics, going from words about the scene taken as whole, to words that apply to specific parts of the scene, to relationships between parts. Methods going beyond keywords are organized by the linguistic roles that are learned, exploited, or generated. These include proper nouns, adjectives, spatial and comparative prepositions, and verbs. More recent developments in dealing with sequential structure include automated captioning of scenes and video, alignment of video and text, and automated answering of questions about scenes depicted in images.
Table of Contents
The Semantics of Images and Associated Text
Sources of Data for Linking Visual and Linguistic Information
Extracting and Representing Visual Information
Text and Speech Processing
Modeling Images and Keywords
Beyond Simple Nouns
About the Author(s)Kobus Barnard
, University of Arizona
Kobus Barnard is a professor of computer science at the University of Arizona. He also has appointments in the School of Information: Science, Technology, and Arts (SISTA), Statistics, Cognitive Science, Electrical and Computer Engineering (ECE), and the BIO5 Institute. He leads the Interdisciplinary Visual Intelligence Laboratory (IVILAB.org). Professor Barnard received his Ph.D. in computer science in 2000 from Simon Fraser University (SFU) in the area of computational color constancy, where his dissertation received the Governor General gold medal awarded across all disciplines. He then spent two years at the University of California at Berkeley as a postdoctoral researcher working on modeling the joint statistics of images and associated text, followed by moving to the University of Arizona. His current research addresses problems in interdisciplinary computational intelligence by developing top-down statistical models that are predictive, semantic, and explanatory. Application domains include computer vision, multimedia data, biological structure and processes, astronomy, and human social interaction. His work has been funded by multiple grants from NSF including a CAREER award, DARPA, ONR, ARBC (Arizona Biomedical Commission), and the University of Arizona BIO5 Institute.