Return to search

A computational model of visual interpretation

Thesis (S.M.)--Massachusetts Institute of Technology, Dept. of Architecture, 2006. / Includes bibliographical references (leaves 48-50). / From the very early phases of design conception, designers use sketches as a powerful design tool. Sketches are however ambiguous. Meanings are associated on fly as the designer 'comes up' with certain ideas while working with it. There is no hierarchy in a sketch. In fact 'structure' is established only after meanings are applied to the sketch. However even in such structurally and conceptually fluid territory, the designer solves most of his design problems and very often comes to quick resolutions. To understand this fascinating tool and how the designer interacts with it, we need to understand how we visually interpret sketches. The process of design is also a reflective act. The designer keeps changing his perspectives and focus based on the unexpected opportunities that emerge from such reflection. Computational systems used today in design exploration are not capable of doing so. On the contrary these systems model the world in a very rigid structured way and cannot produce design ideas beyond what their preset description anticipates. / (cont.) From this perspective there is no novelty, or surprise, in such systems. As a step forward, this thesis proposes the following: 1. Visual Schemas as procedural units of visual memory. They schematically store real world knowledge (courtyard) and form the basis for interpretation. 2. Separation of Shape and Visual Concepts. This thesis suggests that shapes are flat and abstract collection of parts, while visual concepts are subjective and hierarchic ideas, which are formed from the shapes through interpretation. A LISP machine is presented as a basic computational framework for implementing and establishing the model that is proposed. It observes a relatively simple architectural sketch, interprets it reflectively through the activation of potential, alternative contexts, and then gives a collection of concepts that it manages to 'see' in the sketch. / by Kaustuv Kanti de Biswas. / S.M.

Identiferoai:union.ndltd.org:MIT/oai:dspace.mit.edu:1721.1/35129
Date January 2006
CreatorsDe Biswas, Kaustuv Kanti
ContributorsTerry Knight., Massachusetts Institute of Technology. Dept. of Architecture., Massachusetts Institute of Technology. Dept. of Architecture.
PublisherMassachusetts Institute of Technology
Source SetsM.I.T. Theses and Dissertation
LanguageEnglish
Detected LanguageEnglish
TypeThesis
Format50 leaves, 1644290 bytes, 1644883 bytes, application/pdf, application/pdf, application/pdf
RightsM.I.T. theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission. See provided URL for inquiries about permission., http://dspace.mit.edu/handle/1721.1/7582

Page generated in 0.0021 seconds