Euroshop until you drop
Crayon promoted AI-based retail services at the world’s biggest retail trade fair Euroshop2023 in Düsseldorf, Germany — a short retrospective.
What is Euroshop?
Held every three years in Düsseldorf, Euroshop is the world’s biggest retail trade fair where exhibitors offer their products and solutions for the retail industry. Between 22–26 February 2023, an audience of 83 000 people visited Euroshop2023, which covered many dimensions of retail shopping such as Materials & Surfaces, Lighting, Food Service Equipment and Refrigeration & Energy management.
The “retail technology” dimension of Euroshop2023 covered mobile solutions, the latest developments in self-checkout and self-scanning, trends in analytics, RFID, IoT, smart stores and payment, and the latest security technology. People from Crayon Center of Excellence in Vienna and Crayon Germany had the opportunity to present and showcase AI-based retail services at a booth in the “Microsoft Cloud for Retail” area of the retail technology hall — one of a total of 15 exhibition halls.
Attention Economy: Crayon demonstrations
With Crayon offering professional services rather than off-the-shelf products for the retail industry, attention of passers-by for Crayon in the sea of colorful displays was mainly attracted via interactive demonstrations.
The clothes search engine is an interactive demonstration to provide recommendations for fashion items based on visual and textual search. The solution is based on Contrastive Language-Image Pre-training (CLIP) — — an efficient method of image representation learning from natural language supervision. CLIP jointly trains an image encoder and a text encoder to predict the correct pairings of a batch of (image, text) training examples, offering novel options for combined search. The following image shows a screenshot of the demonstration for a pure textual search for flowers.
The next image shows a screenshot of the demo for a pure image-based search for clothes similar to a query image.
The combined image-text search based on both a query image and a text is illustrated in the following image, showing clothes with flowers that are visually similar to the query image.
The influence weight of image search versus text search can be adjusted, offering a new shopping experience. The demo was built on the Zalando Dataset comprising ~8500 image text pairs. More detailed information about this search engine can be found on Julian Pegoraro’s Crayon Tech Blog.
The customer churn prediction demonstration provides intuitive and actionable insights on customer churn, which is the percentage of customers that stopped using a company’s product or service during a certain time frame. Based on existing customer profiles and behavioral data, Crayon’s customer churn prediction provides a deep understanding of what drives churn and opens up possibilities for improvement. The demonstration is implemented in Microsoft Power BI — an interactive data visualization software product developed by Microsoft with a primary focus on business intelligence. The following images show screenshots of three interactive pages of the customer churn prediction demonstration.
The anomaly detection demonstration illustrated predictive maintenance for time series data in order to identify abnormal states or events in a timely manner. Time series data may be, for example, sensors from a 3D printer, a production machine, but also measured indicators related to customer behavior. The following image shows a screenshot of the running demo for three measurement time series, with red areas indicating abnormal behavior, and a general warning indicator about critical behavior.
Anomaly detection applied to image data is a part of Crayon’s offering for Visual Quality Inspection. Quality inspection as part of quality management is mandatory and important for every industry. The quality level of production degrades by defects, which can be automatically detected in imaging data. The detection of defects is based on image examples for good and bad quality. The following image of the anomaly detector demonstration shows the output of an example of a fabric with a defect. On the right-hand side, the heatmap of the defect probability is superimposed on the fabric image.
The Düsseldorf Experience
Conventions and scientific conferences tend to be hosted at touristically attractive locations, and have sometimes the connotation of a ‘vacation’. Now there seems to be common agreement — also among Germans — that Düsseldorf is a far cry from being an attractive place. Nevertheless, after having stood at the exhibition booth a whole day talking to potential customers — Euroshop until you drop — the dinners with Crayon colleagues in ethnic restaurants were entertaining, offering the opportunity for socializing.
Transportation was a hassle — while the strike at Dusseldorf airport on Monday did not directly affect Crayonites from Vienna, the big strike of public transport on Tuesday and Wednesday effectively narrowed down options to uber-expensive Ubers, 90 minutes of walking, or getting a lift from colleagues of Crayon Germany living in the vicinity. Reaching a classical taxi dispatch via a phone call was not possible, of course.
Bottom Line
During the five days of the world’s biggest retail trade fair Euroshop2023 in Düsseldorf, five people from Crayon’s Center of Excellence Vienna and five people from Crayon Germany pitched AI-based retail services, covering fields such as an innovative clothes search engine, customer churn prediction and anomaly detection and visual quality inspection, generating approximately 80 leads.
