- Susanne Bach-Bernhard (for administrative matters)
Recently, psychologists have turned their attention to the study of cast shadows and demonstrated that the human perceptual system values information from shadows very highly in the perception of spatial qualities, in synergy or in conflict with other cues. At the same time artificial vision systems treat cast shadows almost exclusively not as signal but as noise. The purpose of this seminar is to bring together researchers from the various disciplines involved in investigating the problem of understanding the perception of shadows (both in biological and in artificial systems) and experts and practitioners that try to bridge the gap between the perception and the use of the knowledge content in shadows in robotics and computer vision systems.
During the seminar we intend to discuss the ways in which the perception of shadows operates, considering evidence from computer graphics, psychology, art history, philosophy and neuroscience. We shall address the issue of how the environmental information used by the human perceptual system can be incorporated into computer vision methods for shadow detection. Specific issues such as the recognition of shadows ("what makes a dark patch in a scene shadow-like? ") and the human ability to make complicated judgments about 3D location in space based upon shadows are examples of the topics to be addressed.
The seminar has the goal of synchronizing the work of the different research and application communities involved. To that effect, it will provide up-to-date information about the state of the art of research on the psychology of shadow perception, on the practices for depicting shadows, on uses of shadows.
The seminar will also be an excellent forum for researchers in robotics to present the major challenges they have been facing in exploiting the information content in shadows — features that are largely considered noise rather than knowledge in other fields.
The seminar "The Message in the Shadow: Noise or Knowledge?" brought together researchers from the various disciplines involved in investigating the problem of understanding the perception of shadows (both in biological and in artificial systems) as well as art historians and artists involved in the study or in the manipulation of shadows in art pieces. The nationalities of the seminar participants were as varied as the disciplines involved its central theme; from the 20 attendees there were 4 that came from Brazil, 4 from Germany, 1 from the Emirates, 2 from France, 2 from the UK, 1 from Canada, 3 from the US, 1 from the Netherlands and 2 from Japan.
The small size of the seminar helped to create a friendly atmosphere, in which every participant had time and space to engage in discussions with every other, and every one had an equal amount of time to present his/her ideas, independently of the career stage the participant was in.
The dynamics of the seminar was as follows: every participant that had an interest in presenting a talk was allocated a 20 minute slot, followed by a 10 min discussion period, during the mornings (from 9 to 11am). The talks where distributed into 4 tracks, one for each day of the week: Psychology (Monday), Artificial Intelligence and Computer Vision (Tuesday), Art and Rendering (Wednesday), Architecture and Spatial Reasoning (Thursday). The titles of the talks given, per track, are cited as follows (the related abstracts are listed in the next section):
- Patrick Cavanagh. What does vision know about shadows?
- John Kennedy, Shape-from-shadow polarity
- John O’Dea, Do shadows make surfaces look dark?
- Marteen Wijntjes, Perception of shadows in paintings
Artificial Intelligence and Computer Vision
- Hannah M. Dee, Why does computer vision find shadows so problematic?
- Paulo E. Santos, Shadows in AI and Robotics
- Frederick Fol Leymarie, On medialness-based shape representation: recent developments and food for thought
- Ann Marie Raynal, Leveraging the Information in the Shadows of Synthetic Aperture Radar
Art and Rendering
- Koichi Toyama, The systematic introduction of Chiaroscuro in 15th century Florence and the symbolic shadow in Sienese Painting
- William Sharpe, Shadow Messages in the arts
- Marcos Danhoni, Shadows on the moon and the sun by Cigoliand Galileo: The Copernican planetarium inside the Paolina's Chapel of Santa Maria Maggiore
- Roberto Casati, X-From-Shadow: There is still room at the bottom
- Koichi Toyama, Un-naturalistic painting and the lack of shadow: History of shadow in 18th- 19th century Japanese paintings and woodblock prints
Architecture and Spatial Reasoning
- Barbara Tversky, Can uses of shadows in language and art inform perception of shadows?
- Juliano Beraldo, Daylight metrics for building design
- Christian Freksa, Shadow and friends illuminate space
- Mehul Bhatt, Carl Schultz and Jakob Suchan, Grasping Objectified Shadows
At the end of the morning sessions, discussions were conducted in which the ideas presented during the talks served as inspiration for the conception of research statements. Some of these statements were selected to be discussed during the break out session that occurred during the Monday and Tuesday afternoons. The main questions discussed are presented below:
- Information about the light-source contained in shadows: there is a number of features from the light source that is present in the shadow of an object (for instance: the number of sources, the localisation, the shape) but much of this information is not used by the perceptual system. The question of the evolutionary advantages of this selective use of the information content of shadows was discussed and also the possibilities for a computer system to explore it fully;
- Mooney Faces and Shadows: To test people’s vision, Craig Mooney devised two-tone pictures of faces. In Mooney faces, some parts are strongly illuminated, others are in deep shadow. His pictures were static. Motion helps vision find the faces. Mooney faces in negatives are hard to make out. Proper facial expression is lost. In outline, they are equally uninterpretable. Adding a dark line to the border of a positive Mooney face can drop recognition to the level of a negative. Motion helps, but still leaves the face looking cartoonish and flat. Often the line is taken as part of a profile. A light line border of a negative also leaves it cartoonish.
- Cross-disciplinary terminology for shadows: there is currently a non-consensual use of terms to refer to shadow issues (for instance, a caster is sometimes referred to as ‘obtruder’ or ‘occluder’). This group proposed a tentative terminology that was later discussed with the other participants. Throwing away information. Shadows are used by the visual system to retrieve various spatial features of the scene, then discarded. The group discussed cognitive/computational mechanisms that may throw away shadows.
- Mereotopological formalisation of Eclipses. The group created a formalized version of the terminology used in describing the different phases of an Eclipse of the Sun. An amendment of the existing taxonomy was proposed.
At the end of the Monday session, artist Francesca Bizzarri showed some aspects of the art of shadow performance.
On Thursday afternoon the participants were directed to discuss possible collaborations, project proposals, and to devise conclusions (even if partial) to the various questions discussed during the previous days. Some of the results obtained in this session are listed below:
- Collaboration between S. Paulo and Bremen
- Online, real-time, Mooney face generator – A computer generated video by Dee, Kennedy and Casati, on the impairment of depth perception through the display of lines on moving Mooney faces, has been created and is visible at: https://www.youtube.com/watch?v=IuDNUz9RSuw
- Collaboration between Tokyo and New York (on art history)
- The foundation of a work group on terminology
- The projected publication of the mereotopological formalisation of Eclipses (Paris-Bremen-S. Paulo)
- Video displaying the phenomenon of the polarization of shadow (Casati and Cavanagh)
Finally, we discussed the future submission of a proposal for a special issue of the Journal Spatial Cognition and Computation (http://www.tandfonline.com/toc/hscc20/current) with the themes of the seminar and the organisation of a follow-up event in 2017 related to these ideas. Our proposal for a special issue of the Journal Spatial Cognition and Computation was accepted by the journal editors in June, 2015 (the call for papers will be advertised in the second half of this year).
- Juliano Beraldo (University Sao Paulo, BR)
- Mehul Bhatt (Universität Bremen, DE) [dblp]
- Francesca Bizzarri (Associazione Ca' Luogo D'Arte - Gattatico, IT)
- Roberto Casati (ENS - Paris, FR) [dblp]
- Patrick Cavanagh (Paris Descartes University, FR) [dblp]
- Marcos Cesar Danhoni Neves (State University of Maringá, BR) [dblp]
- Hannah Dee (Aberystwyth University, GB) [dblp]
- Valquiria Fenelon Pereina (University Center of FEI - Sao Paolo, BR) [dblp]
- Christian Freksa (Universität Bremen, DE) [dblp]
- John Kennedy (University of Toronto, CA) [dblp]
- Frederic Fol Leymarie (University of London/Goldsmiths, GB) [dblp]
- John O'Dea (University of Tokyo, JP)
- Ann Marie Raynal (Sandia National Labs - Albuquerque, US)
- Paulo E. Santos (University Center of FEI - Sao Paolo, BR) [dblp]
- Carl Schultz (Universität Münster, DE) [dblp]
- William Sharpe (Columbia University, US) [dblp]
- Jakob Suchan (Universität Bremen, DE) [dblp]
- Koichi Toyama (Keio University, JP) [dblp]
- Barbara Tversky (Stanford University, US) [dblp]
- Maarten Wijntjes (TU Delft, NL) [dblp]
- artificial intelligence / robotics
- computer graphics / computer vision
- Computer vision
- Cognitive vision
- Art and cognition
- Spatiotemporal logic and computer vision
- Conceptual representations
- Embedded computer vision systems
- Machine learning and vision