Computer Vision for Fashion: Virtual Dressing Rooms
by Adam Pease
Recently, the computer vision provider, Revery.ai, has turned heads with its exposition of an AI-based solution for virtual dressing rooms. Revery.ai’s press attention has come from the scale and extensibility of its solution, which is capable of analyzing over a million garment images a week. This blog describes the product and considers what it may imply about the relationship between computer vision and generative content.
Who Is Revery.ai?
Revery secured funding last summer from Y Combinator, a seed money startup incubator that helped launch the likes of Airbnb, Stripe, Dropbox, DoorDash, and Instacart. It received backing to the tune of $125,000, with an eventual goal of another $1.5 million seed round down the road. Its small team includes a group of computer science Ph.D. students whose expertise in AI development has led them to create a fast and scalable solution for fashion e-commerce brands.
Revery.ai’s virtual dressing room works by using computer vision to synthetically generate the bodies of human models. It then takes these model images and combines them with databases of e-commerce stock, creating readymade versions of models wearing clothes without having to spend time photographing real human beings.
The End of Models? E-Commerce Goes Generative
In our recently predicted note on 2021 computer vision trends, Aragon predicted the rise in generative solutions to different e-commerce content marketing and advertising challenges. The rise of generative models, such as generative adversarial networks (GANs) has increased as algorithms have become better at producing content that the human eye cannot discriminate as fake.
In the case of human bodies, the future of generative content is already here. Now, brands can feature human faces that have never existed before in their advertisements. This can save a considerable amount of business time and spend when one considers the expenditure involved with designing sets, setting up cameras, editing photographs, and the actual payments to the models. If these solutions prove to be scalable for large brands, we may be heading towards a future for e-commerce that shifts away from reliance on human models altogether.
Revery.ai is just one provider in an e-commerce market that is used to relying on human beings for fashion modeling and content marketing. At the same time, its early success paints an interesting picture of what the future of e-commerce may look like. If businesses can get by generating models for little cost, why would they turn back to real humans?