Mr Lionel Wong
Approaches to applying Machine Learning (ML) to Facilitate Computational Fashion Design
Computational Creativity, Fashion Design, Product Design, Artificial Intelligence and Machine Learning
The advantages of the use of computational techniques in design are numerous. The predictive capabilities of ML allows computers to generate multitudes of possible designs in a defined solution space, which frees up the designer for him to take on a curatorial role, guiding the production of solution outputs. By being a toolmaker, a ‘digerati’ (Oxman 2006), the designer is not constrained by the limitations of the tools she uses, but is able to create new tools to serve specific needs. As computational design lessens the designer’s workload by automating the tedious aspects of solution creation, the designer can harness her liberated energy for deeper creative considerations.
In “The Law of Accelerating Returns”, Ray Kurzwell (2001) describes technological progression as exponential as opposed to linear. He highlights the fallacy of human nature to assume future rates of progress to follow what is in the present, when in reality it is human nature adapting to an accelerating change of pace. It is a common misconception among design practitioners to assume their skillsets will remain relevant throughout their working lives when the prudent mindset is to assume what was learnt at the start of school will be out of date by graduation. As the world moves into the age of Industry 4.0 (Schwab 2016) and beyond, classical designers may need to re-evaluate their skillsets in contrast with design thinkers and computational designers. If the former aims to avoid being edged out by irrelevance, perhaps it behoves these designers, especially those in the fashion industry, to adopt computational design into their toolkits now.
This study will adopt a practice-led approach, where it is defined as being ‘concerned with the nature of practice and leads to new knowledge that has operational significance for that practice’. (Candy 2006) The act of practice, either of the researcher or the research participant, is a definitive part of the fashion design process and therefore it is improvident to separate practice from theoretical research.
Mixed method applications are particularly important when conducted in the context of design research due to the subjective nature of the subject existing alongside the requirement for statistical robustness for the development of professional practices. To fruitfully identify the knowledge gaps of designers in machine learning, we have to be able to capture the full gamut of expression in their feedback while they engage with this new, unfamiliar material. Thus, our research method will allow for qualitative input from respondents such as in the form of verbal elaborations, sketches and self-led projects. And yet, the responses of research participants should be expected to fulfil certain empirical expectations. Relying on one paradigm (qualitative or quantitative) tends to introduce limitations in methodology and derive biased results (Greene et al., 1989) and which we need to introduce triangulation to mitigate such problems. This is done through the corroboration of quantitative and qualitative data obtained from different, consecutive tests for richer data and prompting new modes of thinking. (Rossman & Wilson, 1985)
Results / Outcomes:
- Through a comprehensive literature review on existing trends and approaches, potential applications in the creating and manipulation of 2D and 3D design assets will be consolidated, and the strengths and weaknesses of ML will be contrasted with traditional design methods.
- To understand the knowledge gaps faced by classical designers in their transition to being computational designers and to analyse the stages of learning necessary to reach a level of competency.
- Create a set of sample tools, either in the form of learning frameworks or digital applications, to exhibit the potential of using ML in computational fashion, as well as presenting problems and shortfalls of these new tools to suggest principles for creating ML-driven computational design tools for future development.
Recipient of the Hong Kong PhD Fellowship
Master of Design (Interaction Design) with Distinction
Bachelor of Arts (Industrial Design)
Professor Clifford Choy (Chief Supervisor)
Professor Huaxin Wei
Professor Stephen Wang
Specialization / Interests:
Computational Creativity, Fashion Design, Product Design, Artificial Intelligence, Additive Manufacturing, Sustainability
Expected Date of Completion