Ever since jeweler Louis-Francois Cartier opened his first workshop in Paris in 1847, the identify “Cartier” has been synonymous with distinctive high quality. Nearly two centuries later, the posh Maison has popularized wristwatches for males, been dubbed the “jeweler of kings and king of jewelers” by King Edward VII, and continues to design, manufacture, and promote jewellery and watchmaking creations globally famend for timeless design.
Whereas Maison Cartier prides itself on the vastness of its assortment, manually looking its catalog to seek out particular fashions, or evaluating a number of fashions directly, might typically take fairly a while for a gross sales affiliate at one of many Maison’s 265 boutiques. This was not supreme for a model that’s recognized for its swift and environment friendly consumer service. Thus, in 2020, Cartier turned to Google Cloud and its superior AI and machine studying capabilities for an answer.
Thomas Meyer, Information Officer at Cartier: “We aim to build a global data & digital platform for our Maison, providing our teams with at scale analytics, versatile applications and artificial intelligence capabilities. Google Cloud is to be the core component of this data & digital platform.”
No time to spare: Overcoming product search challenges with AI
Cartier’s purpose was to develop an utility which, when proven a picture of any Cartier watch ever designed in its 174-year historical past, might retrieve detailed details about its particular mannequin and counsel similar-looking watches (with presumably completely different traits equivalent to value) in beneath two seconds. Utilizing the app, gross sales associates would be capable to discover particular merchandise swiftly inside a list of greater than a thousand watches—even some with solely very slight variations between them.
However creating this app meant Cartier’s knowledge staff needed to overcome some distinctive challenges. Coaching machine studying fashions requires enormous volumes of coaching knowledge—on this case, pictures of Cartier wristwatches—however Cartier has all the time been pushed by unique design and its prestigious collections had only a few in-store product pictures obtainable, with variations in backgrounds, lighting, high quality, and styling, making it troublesome to categorize pictures. In consequence, Cartier needed to discover a approach to develop a picture recognition system that was performing above the required benchmarks, because the Maison has very excessive requirements for its consumer service. For the app to be efficiently adopted in its shops, Cartier required no less than a 90% accuracy charge, with the complete pipeline working inside 5 seconds end-to-end when built-in.
That’s when, leveraging their present partnership with Google Cloud, Cartier’s knowledge staff reached out to us for help in search of extra superior machine studying capabilities to carry their imaginative and prescient to life.
Transfer past buyer expertise
Working along with Cartier’s knowledge staff, we had been in a position to assist them remedy their knowledge and visible search problem by means of making an attempt out a variety of machine studying experiments with Google Cloud AI Platform companies, equivalent to AutoML Vision, and Vision API, earlier than rewriting customized code. In the long run we constructed a knowledge mannequin particularly designed for Cartier’s use case: a mixture of classifiers that run in parallel that first acknowledge a watch’s colours and supplies, after which identifies which watch assortment it belongs to. In the long run, it gives a high three record of attainable identities (visually related watches) for the picture, which customers can click on on to get extra data, with as much as 96.5% accuracy inside three seconds.