Rethink Spolia – 2nd Place for Katharina Koska at Heinz-Stillger-Award 2020
The research module “Rethinking Spolia” by Katharina Kostka focuses on circular economy in architecture, by looking back into the past, to find answers for the future. If we reuse components in the future in order to save energy and resources, industrially and serially produced components will no longer be the cheapest and most preferable building material. Instead, demolition material and dismantled elements will be used more often. This paradigm shift is not only an ecological, economic, political and logistical challenge, but will also have an impact on architectural design. What do buildings from the era of circular economy look like? Katharina Kostka tries to answer this question by tracing the use of spolia in the history of architecture. These pieces, which reappeared in new building contexts to demonstrate power or affiliation, can – according to Katharina Kostka's hypothesis – provide insights and inspiration for the design of tomorrow. In her work she traces spolia examples from 2500 years of architectural history. In doing so, she develops a graphic language that enables a comparison of these very different examples. She identifies design categories that have existed over many centuries and that can put buildings into new relationships to each other. The work succeeds in showing spolia in a new context and derives design approaches for tomorrow from art historical research.
DeepPattern – 3rd Place for Paul Steggemann at Heinz-Stillger-Award 2020
The work 'DeepPattern' by Paul Steggemann investigates possible applications of machine learning for architects. In his work, he explores the extent to which an algorithm for the classification of basic architectural figures in floor plans can be trained to support architects in their design work. Such algorithms are currently used primarily in image recognition. In architecture, there are currently only few examples that use such algorithms to support design work.
A major challenge is the provision of training material (floor plan drawings) for the classification algorithm to enable learning. Paul Steggemann used a dataset created by guest professor Ruben Lang and his students at the Unit of Design and Building Typologies. He annotated almost 1,000 floor plans with the necessary information for the algorithm. In initial experiments, the algorithm was tested on previously unknown floor plans and the resulting code is freely available via Github.
The graphical representation and the accessible text clearly communicate the principles and goals of his research work even to laymen in the field of AI. The knowledge gained provides a good basis to further deepen these topics at the Department of Architecture and in collaboration with computer science.