Predicting fashion trends, six months ahead, by analyzing social networks images, it is possible, thanks to artificial intelligence solutions, developed by start-ups like Heuritech. Le Boudoir Numérique spoke with Tony Pinville, co-founder of this french company, collaborating with fashion and luxury brands, such as Dior and Louis Vuitton.

By Ludmilla Intravaia

 Le Boudoir Numérique : You co-founded Heuritech in 2013. How would you define your company ?

Tony Pinville, CEO of Heuritech and PhD in Artificial Intelligence : Heuritech is a company expert in artificial intelligence that helps fashion and luxury brands better understand market and products trends, through the eyes of millions of consumers and influencers, by analyzing the social networks images.

What is artificial intelligence ?

Artificial intelligence brings together all the methods for automating intellectual processes. It can be analyzing images, playing chess, classifying e-mails ..., all the cognitive tasks that we can manage to automate. For example, with visual recognition, information can be automatically extracted from an image. Artificial intelligence includes several subsets, such as machine learning, based on statistical approaches that allow the machine to solve a specific problem by itself, based on data that has been transmitted to it. Take the case of a system having to learn to recognize the distinct presence of dogs and cats in photos. We will provide it pictures of dogs and cats, mentioning "This is a dog", "This is a cat". The system will learn to distinguish a cat from a dog without being told how to do it, by trying to predict, by itself, the characteristics of a cat and a dog. It is the same principle of learning for humans, especially children, when their parents show them two or three pictures of dogs and cats and that they are then able, by generalization, to find by themselves what is a cat and what is a dog.

Why does the fashion industry need your artificial intelligence services ?

Fashion is changing. Twenty years ago, fashion trends were made by key characters, such as Anna Wintour (editor-in-chief of US Vogue, AN) or Karl Lagerfeld (artistic director of Chanel, until his death, in February 2019, AN). But at the present time, trends are increasingly conditioned by influencers on social networks, influencers who can be anyone, posting from anywhere in the world. For brands, it has become more and more difficult to understand what is happening in fashion and the trends that will emerge. In this context, we realized that artificial intelligence represented a technological breakthrough that would allow us to bring new innovative possibilities to the textile industry. That's why we founded Heuritech to harness these AI-related innovations, especially in image recognition, a technology that has surged in recent years.

How is visual recognition applicable to fashion ?

Thanks to artificial intelligence, we are able to pinpoint extremely precise things on social networks, almost more precisely than a human being would. Artificial intelligence can automatically understand what's in an image and, therefore, facilitate visual search, applied here, to the luxury and fashion sectors. Specifically, it is possible to identify, on an image, that a person carries a handbag, that this bag is by Louis Vuitton and that it is the Twist model in metallic black leather. We can also get information about the context, the type of clothing worn by the person, top, pants, etc., not to mention their colors and textures.

How is this information useful for fashion brands ?

Brands are very attentive to market trends. They need to know their evolution in the coming months. Based on the analysis of social networks and their influencers, we are able to predict trends six months ahead. For example, by compiling the visual content of social networks, we can detect that, in some images, a lot of people are wearing skirts, that these skirts are leopard print pleated skirts and so, by analyzing these data, predict the emergence of the leopard print pleated skirts trend, within six months. Brands can then adapt to this forecast by offering leopard print pleated skirts, in one of their collections, since they know that a part of the population will soon be interested in this type of product.

Among your customers, you count Dior and Louis Vuitton. What are the expectations of luxury brands, towards your forecasting tools ?

Alongside some fashion brands that strive to sell the maximum of clothing to the maximum number of people, luxury brands have a different positioning because they must preserve their brand DNA, at the risk of losing their desirability in the eyes of customers. Luxury must remain luxury, an exclusive world, which creations are not worn by everyone. That's why we also help them understand who their customers are. Until now, fashion was obsessed with the product, the creation of new products, without necessarily knowing to which customers they were intended. More and more, different questions arise : Who is the customer ? How does he wear the product ? What use does he make of it everyday ? With our solutions, we help luxury brands better understand their customers and better identify the specificities of each market. For example, in China, unlike another country, the product will be worn in such a way, associated with that other category of products, and so on. More attentive to their customers and potentially inspired by them, luxury brands can, therefore, offer products that retain the DNA of the brand, while corresponding to the evolution of their expectations.

So, you support brands in their decision-making process ?

Indeed, having concrete tools, with accurate data, can help them to choose whether they will launch a new product or not, for example. If I want, let's say, launch a new iconic bag, will it work or not? Through our analysis of social networks and influencers, we can measure the desirability of a product, before its actual launch, by figuring out whether influencers appreciate it, if it leads to desirability among other consumers, etc. In short, this approach can predict if my iconic bag will be a hit or not.

In the long term, could brand stylists rely on this type of forecasting tool to imagine their collections ?

 If we offer a decision support tool, we do not see ourselves as designers telling brands what products to make. On the other hand, as our tool is a guide on market trends, it can be used by luxury brands who, anxious to remain avant-garde, would like to position themselves in the opposite of its predictions. Similarly, fashion designers can draw on sources of inspiration and new directions on different uses of products. It is not a tool intended to replace man but rather who will help him to go further, in his process of creation, as any tool should allows it, in fact.

Can we say that artificial intelligence fits into the context of fashion personalization ?

Absolutely. Thanks to all these datas, from social networks, which we did not have before, we tend to a customization, not only focused on individuals, but also on more precise productions. If we have, for instance, identified a niche market for the leopard print pleated skirts, in such region of France, affecting such number of people, it will be easier to produce, in a very reactive way, the precise quantities, in factories close to the sales outlets, in order to respond to this need that would have been difficult to apprehend, in the absence of information about it.

In your opinion, could artificial intelligence promote a more sustainable fashion, leaving less room for waste?

Fashion faces important issues, generated by its non-personalization, namely the problems of lack of production precision and clothing overstocking. For example, H&M made so much clothing last year that it faced $ 4.3 billion worth of unsold products. And this brand is far from being the only one affected by this phenomenon. Many brands destroy their unsold products because they find it difficult to evaluate the quantities to be produced. Understanding better the trends and expectations of customers can defeat this phenomenon, producing, at the right time, better clothes that stick more to consumers desires, limiting as much as possible disappointments at the reception of packages, product returns and the waste of unsold products. While textiles represent the second most polluting industry in the world, technological tools should, more generally, participate in improving the fashion sustainability and the end of ecological waste.

 * Heuritech website is here. 

 * This interview with Tony Pinville took place while doing research for the article entitled "Shopping perso 3.0", for the Belgian magazine Le Vif Weekend, dated March 21, 2019. Read it here.