Genostyle x LCF
Genostyle x LCF
The Fashion Innovation Agency recently supervised post-graduate students from London College of Fashion, working with Genostyle, on a cutting-edge project that used Artificial Intelligence (AI) to measure fashion style.
Genostyle’s technology enables individual customers’ preferences and styles to be characterised and tracked by computer on a massive scale, making possible customisation and online personalisation via advanced algorithms and AI. The need for this capability is well recognised:
The aim of this project was to create a universal, western, fashion style taxonomy by decoding style DNA of brands via algorithms, in order to improve customer service through style recommendations.
Students were split into three different groups (two focused on womenswear and one on menswear). Their objectives were:
– To identify resources for fashion trends/style forecasting
– To generate a list of fashion styles to update Genostyle’s Taxonomy Database. The identified style taxonomy includes: a) Comprehensive styles (covering most ‘western fashion’ styles and b) Relevant styles (reflecting contemporary styles)
– To research and identify the top 10 brands representative of each style. These brands were used for benchmarking and as input in running Genostyle’s algorithms
– To research and identify an additional set of 2 to 3 brands representative for each style. These brands were used to measure the accuracy of the proprietary algorithms in predicting brand style (algorithm testing brands)
– To research and identify influencers representative of each style
– To identify specific items representative of each style and provide ‘mix & match’ looks to represent these fashion styles
– To collate all of their research and present it in a report, clearly demonstrating the fashion styles and respective brands/influencers
– To learn how to work with the latest trend-predicting technologies (AI and style-developing algorithms) and to learn about their application
Using the style taxonomy and brands provided by the student groups, Genostyle ran the proprietary algorithms. The information was used to “teach” the algorithm on how to classify the brands by styles rather than merely shapes and colours, mirroring the thought process that was carried out when initially defining the styles. The purpose of this testing was to find out how much the machine classification was consistent with the manual classification.
The final algorithmic results determined the extent to which the algorithm categorised the style of each brand, demonstrating whether it matched the initial definition/category, or whether it deviated from the suggested classification. The results were defined as a percentage hit rate, which implies how accurately the brand-classifications were successfully tested, verified and categorised by the Genostyle style taxonomy and therefore how well the algorithm has been taught to style and recognise the brands’ styles.
What are the implications of fashion brands using artificial intelligence?
The key shift will see fashion brands switching from a brand-centric supply model to a customer-cetric demand model. AI and machine learning can be used to better define what a consumer may want in the future. This knowledge changes the design process, as instantaneous data defines what is designed now, rather than designing for a far future. Genostyle’s learning algorithm can predict which styles sell best and detect the overall features of these styles to inform and steer a designer’s collection. This efficiency would reduce the number of garments going into markdown and increase profitability. This insight into what consumers want allows brands to understand their customers’ personas, allowing them to offer a heightened level of personalisation and customer service. This understanding of ‘big data’ also enables brands to get the right marketing communication to the desired target group with precision.
The LCF postgraduate students that took part in this Collaborative Unit project were: Nathalie Bacardi, Silvia Mantoni, Yasmina Nessim, Sanghamitra Samaddar, Tekila Nobile, Megan Bolotin, Giulia Wilzewski, Sarula Bao, Vandana Pratap, Wendy Wang, Jinzhi Xu, Yinyu Ding, Yiwei Lai and Yuan Huang.