Venerdì 27 ottobre 2023, alle ore 16, si svolgerà la trentunesima Kanizsa Lecture nell’Aula Magna dell’edificio centrale A del comprensorio universitario di piazzale Europa 1, a Trieste.

Il relatore di quest’anno è Phil Kellman, scienziato dell’Università di Los Angeles, California (UCLA).

Abstract dell’intervento di Phil Kellman "From Fragments to Objects: Understanding Visual Completion in Object Perception":
        
Perceiving objects and their properties is crucial to thought and action. Through vision, our most powerful sense for obtaining knowledge of objects, we are somehow able to perceive complete objects despite fragmentary optical inputs due to occlusion and camouflage. How is this accomplished? I will describe a progression from several decades of work by many researchers to recent research and theory that suggests a coherent, integrated account of the processes that produce perception of complete objects of determinate shape despite spatial gaps in the input. 
This account incorporates well-understood properties of distinguishable and complementary contour and surface interpolation processes, and also fits with a variety of phenomena suggesting additional scene constraints that modulate the strength of visual completion. Persisting controversies about how these fit together may be resolved by distinguishing two processes. An automatic contour-linking process that connects edges based on geometric relations in 2D, 3D, and spatiotemporal inputs produces an intermediate representation that is prior to perceptual experience. A subsequent scene description process sustains, weakens, or even deletes linkages in the intermediate representation based on a number of interacting scene constraints. The outputs of the scene description process determine perceptual experience. I will show experimentally how this approach helps to understand path detection, a perceptual phenomenon that has been poorly understood, and connects it to phenomena of modal and amodal completion. I will also describe recent research that reveals some of the workings of the scene description process. Evidence suggests that scene constraints that modulate perception of interpolated edges combine in a simple additive fashion, but only for linkages encoded in the contour-linking process. These findings indicate the importance of computations on intermediate representations in perception and also illustrate the kinds of evidence that can be used to reveal them. The approach also suggests that general theories of perception emphasizing automated perceptual mechanisms and those invoking more open-ended interactions of multiple constraints are both correct in describing different aspects of object formation.
 

L'evento può essere seguito anche in streaming dal canale YouTube di Ateneo.