Using Computer Vision to predict Fashion Trends
In today’s fashion world, millions of pictures are posted each day on social media, from consumers and influencers. For brands, it’s a tremendous opportunity to tap into current and future trends – with the right tools. At Heuritech, we analyze 3 million images each day on social media and we turn that data into meaningful insights about trend behaviors, demand planning and stock quantities management. Computer vision for fashion is very challenging: the input images from the open world are very diverse, and the nature of the detected attributes (colors, patterns, shapes, products) are constantly evolving with new collections. As the CTO & AI lead at Heuritech, I’ll detail our research and engineering approaches to tackle this problem.
Charles Ollion is the CoFounder and responsible for AI & Computer Vision at Heuritech. He built a team at the crossroad between academia and engineering, with 8 PhDs and 40 total people - focusing on deploying state-of-the-art solutions at scale. Charles Ollion holds a PhD in Machine Learning from Sorbonnes Universités, Paris and is Lecturer in Deep Learning at Ecole Polytechnique's DataScience master.