Recommendation Engines and Visual Search

Understanding consumer behavior has never been more critical for retailers. To drive growth, intelligent recommendations and augmented reality (AR) environments are being used to create tailored experiences. To improve revenue, online retailers are using GPU-powered machine learning (ML) and deep learning (DL) algorithms for faster, more accurate recommendation engines.​ And, AI is now key for the growing trend of buying online and picking up in store (BOPIS).

Recommendation Systems

On some of the largest commercial platforms, recommendations account for as much as 30% of revenue, which can translate into billions of dollars in sales. That’s why retailers are using recommender systems to drive every action shoppers take, from visiting a web page to using social media for shopping. They also improve conversion by offering up relevant consumer products from the exponential number of available options. 

NVIDIA Merlin, an end-to-end recommender-on-GPU framework, provides fast feature engineering and high training throughput to enable fast experimentation and production retraining of DL recommender models. Merlin also enables low latency, high-throughput, production inference.

Personalized Recommendations

To engage consumers, retailers need to deliver on an expectation of one-to-one personalization. Olay Skin Advisor, a GPU-accelerated AI tool that works on any mobile device, assesses a user-provided selfie and recommends an Olay regimen to improve trouble areas. After four weeks, 94% of Skin Advisor users continued to apply the recommended products. 

Stitch Fix, a fashion ecommerce company, is piecing together a seamless balance between AI-powered decision making and human judgement. By using algorithms to understand customer preferences, Stitch Fix created a fashion service that combines the art of personal styling with data analytics—all powered by GPU-accelerated DL.

Product Filtering

Retailers are leveraging the next generation of computer vision for sophisticated image attribute recognition to automatically generate comprehensive meta-tagging and cataloging. Access to comprehensive information about products and services helps identify images, resulting in a successful personalized recommendation system. 

Since fashion changes quickly, NVIDIA partner Omnious offers an AI-tagging API that helps B2B customers stay ahead of the fashion curve. Ominous Tagger, the automated tagging solution with over 95% accuracy, is 100X faster than manual tagging and increases search efficiency by 4X. Omnious also offers a trend report that analyzes social media fashion influencer images.

Virtual Fitting

The 2019 cost of returned merchandise in the US was $309B. Online returns accounted for $41B of that total.  To reduce the number of returns and provide a more enhanced shopping experience, retailers can now suggest items to customers that are virtually guaranteed to fit. 

Cappasity enables customers to experience a virtual fitting to see how garments look on, before they buy, using its 3D Virtual Try-On solution. Powered by NVIDIA GPUs, with CUDA to boost the speed of calculations, Cappasity’s algorithms process data in the cloud to detect body measurements, while neural networks perform human contouring segmentation.

Sign up for the latest retail news from NVIDIA.