Solution brief

AI Machine Vision
Luxury Retail

WonderStore Creates Smart Brand

Detection for Deeper Customer Insights

Real-time customer insights with brand recognition powered by the

Intel® Distribution of OpenVINO™ toolkit

Responsive retail for a relevant customer experience

Staying competitive in fashion retail requires outstanding customer experience for
demanding consumer segments. Understanding customers—what they’re wearing
and how they interact with the retail environment—is key to increasing buy
conversion. Now, WonderStore, powered by Intel® technologies, allows retailers to
“Intel® DevCloud for the Edge collect more data than ever about their customers, with a 16 percent improvement
in shop-window conversion for existing customers.1
allows us to choose our
hardware configuration to The new WonderStore Brand Detector, optimized with the Intel® Distribution of
OpenVINO™ toolkit and Intel® DevCloud for the Edge, identifies the type and
optimize training, enabling us brand of clothing being worn by shoppers. The solution automates brand
to create vertical CV models identification, using visual sensors directed at shop windows, entrances, and store
for each customer’s needs.” shelves. Using WonderStore Brand Detector, luxury retailers can better understand
what customers are wearing and where they spend time in the store, then offer
—Reinier van Kleij, CTO, WonderStore
them relevant recommendations based on existing preferences.

Challenges: Brand identification training and visual

recognition frame rates
Identifying clothing type and brand on a consistent basis offers several challenges for
retailers seeking an automated solution. The first is that customers may be wearing
any of a large number of clothing brands, and it is often difficult to obtain catalog
content images. Even when these images can be sourced, they must be regularly

16
updated as new products are released. Visual identification represents a compute

%
challenge: high frame rates improve the accuracy of automated sentiment and
clothing brand analysis, but maintaining them can create a high computational burden.
In addition to in-store challenges, developers face a time-consuming process to
improvement in shop- first create effective deep learning prototype models, then optimize those models
window conversion for
existing customers1 for maximum performance. However, developing and iterating quickly on these
models can offer a significant competitive advantage for retailers, who can use
them to gain deeper insights about each customer in or around their storefront.

Solution: Optimized brand recognition for improved

store performance
WonderStore gathers data from cameras located throughout the store, in addition to
those in shop windows. Using computer vision algorithms and artificial intelligence
trained on models from the Intel Distribution of OpenVINO toolkit, customer data is
analyzed, aggregated, and anonymized before being sent to the cloud.
Solution Brief | WonderStore Creates Smart Brand Detection for Deeper Customer Insights

In real time, the results of this analysis are used to power Benefits of WonderStore’s audience measurement
a range of smart shopping experiences in store, as well as and brand recognition include:
communicate data with sales associates and retail marketing
teams. This enables rapid A/B testing of shop windows or • Improved shop-window conversion rate: WonderStore
in-store displays, as well as development of visitor personas helps turn curious shoppers into customers, increasing the
to increase revenue and improve targeting. shop-window conversion rate by 16.7 percent by helping
retailers better understand which customers are looking at
WonderStore enables streaming content, personalized by their window displays and for how long.1
visitor segment, to be pushed to store displays. This creates a
dynamic, interactive space that adapts to fit target customers. • Optimized personnel management: Personnel and
With sentiment analysis and dwell-time monitoring, stores workloads can be managed in a data-driven way, with
can better understand which displays are making customers actionable insights about visitor flow, segmented by type
happy—and which brands are being worn by the customers of visitor, on a daily and hourly basis.
who are responding best. • Customized customer interaction: Using visitor
Using advanced computer vision technology, WonderStore can segmentation and real-time profile KPIs plus analytics,
use a previously submitted series of opt-in selfies of a shopper interaction with customers can be made smart, with relevant
(from sources like social media drawings, customer reward messages, music, and even temperature and lighting
programs, or email marketing contests) to identify the shopper optimized to drive maximum sales to the target customer.
in store and determine which online marketing campaigns • Omnichannel experiences: Enriching customer data
are generating results at the retail level. with external applications makes it possible to improve
email targeting and social media engagement efforts with
in-store customers for a seamless transition between
responsive retail stores and personalized marketing.

Use cases: The retail store of the future, today

With real-time recommendations that enhance the shopping experience, plus in-depth analysis and deep customer
insights to enable data-driven retail strategy, WonderStore can create a smart store that truly knows its customers.
The complete WonderStore smart store platform enables a wide range of use cases to create a responsive, relevant
luxury retail experience, including:

Intelligent digital signage Visitor insight Loyalty Cardless service Immersive mirror
Learn how many visitors Better understand what After a customer takes This semitransparent
and which visitor types types of visitors enter the a series of selfies in the mirror with a high-
stay close to or in front of store, as well as their daily WonderStore Loyalty intensity display behind
digital displays, with full and hourly use patterns mobile app, WonderStore it recognizes the clothing
analysis of clothing type and where they spend can activate the Loyalty a store visitor has tried
and brand, dwell time, the most time once inside Cardless service, on and can use clothing
and sentiment, all while the store, for improved recognizing the customer information and other
streaming content relevant performance of each on sight with no card KPIs to make targeted
to them. store area. needed. recommendations to the
customer in real time.

How it works in brief

To reduce training times associated with developing deep Developers can rapidly iterate on new models for inference
learning models, WonderStore tapped into the deep from video or images, with scalable storage that easily
ecosystem of Intel® technologies. WonderStore leveraged accommodates large data sets, and leverage pretrained
the pretrained deep learning models included with the Intel models to quickly develop new solutions, like the Brand
Distribution of OpenVINO toolkit to rapidly improve the Detector, on top of existing ones.
quality and performance of its visual recognition features,
like its head-pose model.

2
Solution Brief | WonderStore Creates Smart Brand Detection for Deeper Customer Insights

WonderStore built its solution with efficiency in mind.

Optimizing deep learning models with the Intel Distribution About WonderStore
of OpenVINO toolkit made it possible for WonderStore to run
WonderStore is an Italian luxury computer vision boutique
these and other models concurrently on an entry-level Intel®
leveraging the Intel® Distribution of OpenVINO™ toolkit
NUC with an Intel® Core™ i3 processor, 16 GB of RAM, Intel®
for acceleration of CV models on Intel®-based hardware in
HD Graphics GPU, Intel® Movidius™ Myriad™ X VPU, and a 128
order to provide in-store analytics of shopper profiles
GB SSD. To maintain an up-to-date catalog of brand images
with attributes like dwell time and brands of clothing and
for brand recognition training, image files were automatically
accessories.
scraped from brand e-commerce sites, but in the future,
high-resolution images will be supplied by the brands. The WonderStore computer vision service helps brick-
and-mortar retailers better understand their customers’
Using the toolkit also enabled WonderStore to transition desires and anticipate shopper behavior. The service
from using a CPU-based algorithm for its visual recognition combines hardware and software components with an
and head-pose extraction model to using a GPU-based intuitive business intelligence dashboard that helps
algorithm running on low-cost Intel HD Graphics. With an visualizes profiles of store visitors, including emotions,
economical PC profile and little programming effort, this clothing brands, and shopper flow.
simple change enabled a dramatic performance boost: from
an average of 15 fps with CPU processing to 60 to 65 fps
using the parallel processing power of the integrated GPU.1

Conclusion: Microtargeted retail made “Because WonderStore creates specific CV models

possible by Intel technology optimized for for specific customers, we need OpenVINO™ to be
deep learning able to run many models smoothly, using affordable
WonderStore created a GDPR-compliant smart store hardware. By leveraging a huge repository of
platform that uses the power of computer vision to segment publicly available models for the CV developer
visitors and better understand their behavior. Powered by
Intel® processors, Intel HD Graphics, Intel® VPUs, the Intel
community offered by the toolkit, WonderStore can
Distribution of OpenVINO toolkit, and Intel DevCloud for the provide in-store analytics and detailed shopper
Edge, the WonderStore platform offers luxury fashion retailers profiles—not just gender and age, but also which
a new way to see customer interaction and purchases. clothing and accessory brands they wear.”
WonderStore uses in-store visual sensors and real- —Reinier van Kleij, CTO, WonderStore
time analysis to customize store experiences based on
customers’ fashion choices, sentiment, and dwell time.
Retailers can also make more-effective decisions about
store displays, promotions, marketing campaigns, and
personnel scheduling, using predictive analytics to guide a
hypersegmented, data-driven retail strategy.

Learn more
To discover how the WonderStore smart store platform can
drive more-precise customer segment and retail store insights,
visit wonderstore.ai today.

1. Source: Internal WonderStore performance data.

Intel does not control or audit third-party data. You should review this content, consult other sources, and confirm whether referenced data are accurate.
Cost-reduction scenarios described are intended as examples of how a given Intel-based product, in the specified circumstances and configurations, may affect future costs and provide cost
savings. Circumstances will vary. Intel does not guarantee any costs or cost reduction.
Intel® technologies’ features and benefits depend on system configuration and may require enabled hardware, software, or service activation. Performance varies depending on system
configuration. No product or component can be absolutely secure. For more complete information about performance and benchmark results, visit intel.com/benchmarks.
Intel is committed to respecting human rights and avoiding complicity in human rights abuses. See Intel’s Global Human Rights Principles. Intel’s products and software are intended only to be
used in applications that do not cause or contribute to a violation of an internationally recognized human right.
© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.
0820/DC/CMD/PDF
3