Amazon has revealed that its new AI model enables iterative refinement of results and better color matching. For instance, a shopper could search on “women’s black pants”, then add the word “petite”, then the word “capri”, and with each new word, the images on-screen would adjust accordingly.
Amazon’s system can help retain old visual features while adding new ones, the company said in a blog post. The other is said to be a color model that yields images whose colors better match the textual inputs.
Generative adversarial networks (GANs), first introduced in 2014, have been successful at generating synthetic images. A GAN consists of two networks, one that tries to produce convincing fakes, and one that tries to distinguish fakes from real examples. The two networks are trained together, and the competition between them can converge quickly on a useful generative model.
In a paper that was accepted to IEEE’s Winter Conference on Applications of Computer Vision, scientists at Amazon describe a new use of GANs to generate examples of clothing that match textual product descriptions. The idea is that a shopper could use a visual guide to refine a text query until it reliably retrieved the product for which she or he was looking, the blog post explained.