Overview | Research | Grants/Awards | Teaching | Publications
As a child, Michael Tarr's father, Joel, a historian, frequently whispered into his young son's ear, "Think science, engineering, and math, not history." This subtle psychological manipulation seems to have had some effect in that Tarr started reading science fiction at age 12, bought a slide rule at 14, began programming in FORTRAN at 15, joined the high school math team at 16, took his first engineering course at 17, never seriously considered a career in history, and ultimately received his Ph.D. in Brain and Cognitive Sciences from MIT. For this early advice Tarr is very thankful to his dad -- although he still hasn't entirely forgiven him for selling his dark blue TR3 convertible when the family moved from Southern California to Pittsburgh in 1967.
Pittsburgh, while not quite as serene or as warm as California, had other virtues: close proximity to Carnegie Mellon University (where Tarr took his first cognitive psychology class and worked for several economists and psychologists), winning sports teams with the likes of Roberto Clemente and Mean Joe Green, and a declining steel industry that rapidly led to the cleansing of Pittsburgh's air and water. Tarr's first brushes with the field of psychology occurred in Pittsburgh. Perhaps the most notable was being mentioned in the introduction of his stepmother Tova's Ph.D. dissertation on adolescence as not exhibiting any of the traits typical of adolescents.
In 1980 Tarr enrolled at Cornell University. Searching for a major, Tarr took a course with Ulric Neisser and decided that cognitive psychology provided a fascinating blend of behavioral studies, artificial intelligence, and neuroscience. Upon returning from a semester abroad in Denmark, Tarr was pleased to discover that the psychology department had hired three young and dynamic cognitive scientists: Kay Bock, Jennifer Freyd, and David Irwin. Tarr was greatly influenced by all of them with Freyd serving as his senior honor's project advisor and Irwin providing a great deal of support and advice. Although Cornell was one of the premier places to study the ecological approach to visual perception, Tarr took one of Irwin's best pieces of advice -- the suggestion that Steve Pinker was one of the most productive young scientists studying high-level vision and that working with him might be interesting. Thus, Tarr left ecological approach and Cornell for MIT, where the computational approach was considered the way to do things.
Irwin's words were prophetic -- Pinker certainly was productive and fun to work with. Fortunately by the time Tarr got to MIT Pinker had moved almost entirely into the study of language, leaving the study of visual object recognition wide open. Before Tarr could tackle this problem, however, he required some conversion regarding the "MIT-way." This was easily accomplished through the efforts of a close-knit cohort of graduate students and an interesting and lively group of faculty. Most importantly, although often deep in thought about the past tense, Pinker was good natured enough to collaborate closely with Tarr on questions in object recognition, mental imagery, and visual perception. In particular, Tarr was deep in thought about how human perceivers are able to visually recognize 3D objects from new viewpoints given that our eyes only receive 2D images. Tarr and Pinker's research on this question led to the development of the "multiple-views" or "view-based" approach to object recognition -- the critical insight being that human perceivers remember objects as collections of "snapshots" that are highly specific to the original viewing conditions and that normalization processes are used to match unfamiliar viewpoints to known views. This basic concept still guides much of Tarr's work to this day.
Soon after completing his Ph.D. on the view-based approach, Tarr accepted a position as an assistant professor in the Department of Psychology at Yale University. Unwittingly Yale also hired two other faculty with interests in high-level vision: David Kriegman in Electrical Engineering and Greg Hager in Computer Science. Kriegman, Hager, and Tarr became fast friends and began interacting on a regular basis (particularly during the no-cost lunches offered to Yale faculty in the student dining halls -- proving that there is such a thing as a free lunch). The three of them founded the Yale Vision Lunch seminar series and began building an interdisciplinary program focusing on understanding vision in both biological and artificial systems. Through their various lunches Tarr gained several pounds and a far better knowledge of state-of-the-art work in computer vision and robotics.
Yale's faculty community was both supportive and stimulating, but it was the arrival of three extraordinary graduate students, Will Hayward, Isabel Gauthier, and Pepper Williams, that sparked many of the successes that led to the Early Career Award. Each of them has contributed a great deal to Tarr's thinking regarding the nature of object recognition. With the three of them present (as well as Isabel's dog, Salvador), the lab became a highly synergistic place in which new ideas and discoveries became almost routine. Another element of this synergy was Scott Yu, a Yale undergraduate who, following the creation of the infamous "Greeble" objects for his senior project stayed on as the lab "graphics guru." Several other Yale undergraduates also helped make the lab an exciting place to work: Steve Messe, Jaymz Rosoff, and Jay Servidea. The training of these and other graduate and undergraduate students and their subsequent successes gives Tarr as much gratification as any of his other academic accomplishments.
Another important event during Tarr's time at Yale was an invitation from Heinrich Bülthoff to visit him at the Max-Planck Institute for Biological Cybernetics in Tübingen, Germany. Bülthoff and Tarr begin a fruitful and enjoyable research collaboration on the types of features that comprise view-based representations. Equally important, Tarr shared an office with Dan Kersten, who proved to be a remarkable scientist and a wonderful colleague. Kersten's probing questions regarding the mechanisms used for recognition and his clear thinking regarding unsolved problems in object vision were immensely helpful to Tarr in shaping his larger view of how object recognition might work. Kersten and Tarr also began to explore other aspects of the visual recognition problem -- for example, how varying illumination affects object recognition performance. Beyond any specific project, Bülthoff's collegiality and willingness to accommodate visitors to his institute facilitated an informal, but highly productive, working group on object vision -- along with Bülthoff and Kersten, colleagues such as Shimon Edelman, Mel Goodale, Tommy Poggio, Philippe Schyns, Shimon Ullman, and Alan Yuille have all helped to improve the quality and the breadth of Tarr's work.
Since arriving at the Department of Cognitive and Linguistic Sciences at Brown University, Tarr has moved in a variety of new research directions in cognitive neuroscience -- an area that has great potential for helping us understand the functional architecture of cognition. Through the generosity of John Gore and Adam Anderson he and Gauthier have been using functional Magnetic Resonance Imaging to explore which brain areas are active during the recognition of objects and faces; through the generosity of Marlene Behrmann they have also been studying the specific visual recognition deficits of brain-injured individuals suffering from object agnosias (impaired face or object recognition). Tarr has also benefited greatly from Brown's uniquely interdisciplinary community of scholars studying computation, the mind, and the brain. He is currently involved in developing new collaborations that will use neural-networks to model visual recognition and virtual reality to study visual navigation and scene recognition (with an adherent of the ecological approach no less).
Although he has never published a paper with her, the most important collaborator throughout Tarr's career has been his wife, Laurie Heller. Not only does Tarr depend on Heller's personal support and encouragement, but he frequently relies on her intellectual help to keep him honest with regard to psychophysical methods and his sometimes wild theorizing about the nature of human object vision.
Download Michael Tarr's Curriculum Vitae in PDF Format
