High-Level Vision: Recognizing Objects and Natural Categories

Shimon Ullman

Weizmann Institute of Science, Israel

 

Brief description:

The study of high-level vision is aimed at understanding how visual perception is used to understand the world around us, in terms of objects, categories, events, agents, actions, goals, social interactions, and the like. In this series of presentations I will examine problems of high-level vision with focus on the recognition of objects and natural categories. Specific topics will include: the recognition of individual objects and natural categories, feature hierarchies in the visual system and their acquisition, figure-ground segmentation, bottom-up and top-down processing, innate mechanisms and learning, and neuronal aspects of visual awareness. I will address these issues primarily from a computational perspective, but will include some related material from brain and cognitive studies.

 

Session 1:  Learning to recognize objects and natural classes.

Readings:

S. Ullman, M. Vidal-Naquet and E. Sali, Visual features of intermediate complexity and their use in classification. Nature Neuroscience, 5(7) 1-6, 2002.

Supplementary:

Biederman, I. (1987) Recognition-by-components: a theory of human image understanding. Psychol Rev. 94(2), 115-47. 1987

Burges, CJC.  A Tutorial on Support Vector Machines for Pattern Recognition. Data Mining and Knowledge Discovery, 2, 121–167 (1998)

 

Session 2:  Feature hierarchies and the visual system.

Readings:

Epshtein, B. Lifshitz, I. Ullman, S. Image interpretation by a single bottom-up top-down cycle. Proc. Nat. Acad. Sci., 105(38) 14298-14303, 2008.

Supplementary

Kschischang FR, Frey BJ, Loeliger HA (2001) Factor graphs and the sum-product

algorithm. IEEE Proc Info Theor 47:498–519.

 

Session 3: Combining recognition and segmentation

Readings:

Ullman, S. Object recognition and segmentation by a fragment-based hierarchy. Trends  in Cognitive Sciences (11)2, 58-64, 2007

Supplementary

Borenstein, E. and Ullman, S. Combined bottom-up and top-down segmentation. IEEE PAMI 30(12), 1-17, 2008. 

 

Session 4: Beyond classification: new directions in visual cognition  

Supplementary:

Human cognition and brain:

Elizabeth S. Spelke and Katherine D. Kinzler. Core Knowledge. Developmental Science 10:1  89-96 (2007). 

György Gergely, Harold Bekkering, Ildikó Király. Rational imitation in preverbal infants. Nature, 415, 755 (2002).

Roelfsema, P.R. Elemental operations in vision. TRENDS in Cognitive Sciences 9 (5) 226-233, 2005.

Computational:

Ullman, S. Visual routines. Cognition, 18, 97-159, 1984.

Karlinsky,L.  Dinershtein, M. Harrari, D. Ullman, S. The chains model for detecting parts by their context. Accepted to: Computer Vision and Pattern Recognition (CVPR) 2010.

 

Session 5.  On visual awareness and the brain.

Supplementary

Tong F, Nakayama K, Vaughan JT, Kanwisher N. Binocular Rivalry and Visual Awareness in Human Extrastriate Cortex. Neuron, 21, 753–759, 1998.

Stoerig, P. and Cowey, A.  Blindsight in man and monkey. Brain 120, 535–559, 1997.

Amedi A, Stern WM, Camprodon JA, Bermpohl F, Merabet L, Rotman S, Hemond C, Meijer P, Pascual-Leone A.  Shape conveyed by visualto-auditory sensory substitution activates the lateral occipital complex. Nat Neurosci. 10(6), 687-689, 2007. 

 

Shimon Ullman

Shimon Ullman is the Samy and Ruth Cohn Professor of Computer Science in the department of computer science and applied Mathematics, the Weizmann Institute of Science, Rehovot, Israel. Prior to the current position, he was a Professor at MIT in the Brain and Cognitive Science and the Artificial Intelligence Laboratory. His main areas of research are vision and brain modeling, with emphasis on object classification and recognition. He is the author of  the books: The Interpretation of Visual Motion (MIT Press, 1979) and High-level Vision (MIT Press, 1996). He is the recipient of the 2008 David E. Rumelhart award, and the Hilgard Scholar, Stanford University, 2009