Retail-focused computer vision Foundation Model identifies hundreds of thousands of consumer goods with comparable accuracy to humans during in-store pilot
UltronAI announced the availability of the UltronAI platform, an AI-based computer vision engine built to power the next generation of product-centric retail solutions. With the general availability of the platform, UltronAI has begun to raise its seed round of funding, which follows the successful close of its previous, oversubscribed pre-seed round.
Designed explicitly to work in a high-volume, transactional environment, UltronAI harnesses innovation from decades of research to address unique retail challenges such as high-speed transactions, enormous (and constantly evolving) product catalogs, and variabilities in lighting and product placement. Retailers and retail solution providers can now use the UltronAI platform to develop scalable solutions around loss prevention, self-checkout, inventory, analytics, and more.
Product identification built for retail
As more retailers seek to capitalize on the promise of computer vision, widespread adoption in retail has been hindered by limitations on product catalog size, inconsistent store lighting conditions, user behavior, transactional speed, and cost. The UltronAI platform was built explicitly to overcome these challenges in the retail sector.
Specifically, UltronAI’s technology has been calibrated to:
- Support large and growing product catalogs: In addition to the hundreds of thousands of consumer products UltronAI can recognize today, its zero-shot enrollment enables retailers to add new products in just seconds from a single catalog image, rather than the hours or days required by other solutions. In recent tests, UltronAI ingested a large retailer’s 250,000+ product catalog in less than 45 minutes.
- Work in real-world store conditions: Built to achieve accuracy despite environment conditions and without user training, UltronAI excels at identification even when lighting is low and/or products are obstructed or positioned haphazardly.
- Achieve fast, accurate identification: In a real-world, in-store pilot deployment by a large U.S. retailer, UltronAI achieved accuracy comparable to a human against the store’s actual product catalog.
- Simplify deployment: Using an SDK and API-first architecture, UltronAI has been designed to speed innovation and time-to-value. This is further reinforced by the ability to support both cloud-based and edge-based architecture models.
Read More: GreyOrange Raises Item-Level Accuracy to 99% with gStore Overhead RFID Technology
“The biggest challenge to adoption for computer vision-based product identification is taking it from the lab in small-scale tests to the streets in very large-scale database galleries,” said Dr. Marios Savvides, Founder, Chairman, and CTO of UltronAI. “It’s quite easy to recognize a product from a small gallery when positioned perfectly under bright lights. But correctly identifying an obscured product under poor lighting out of a database of hundreds of thousands of products is actually quite difficult. UltronAI is the outcome of two decades of AI research around object identification and robust, large-scale face recognition. We’ve purpose-built a platform that can support the volume, speed, and store conditions that a typical retailer needs to support.”
Empowering a new generation of retail solutions
UltronAI’s ability to perform accurately in large, complex, and fast-paced retail environments means that retailers and retail solution providers can envision scalable solutions to meet the retail industry’s most challenging problems. Previously, such innovation might be hampered by the limitations of the computer vision engine, UltronAI can fuel new approaches to:
- Loss prevention. UltronAI’s embeddable engine can detect and identify products, validating purchased items and making sure that none get missed.
- Self-checkout. Simultaneous product identification can improve checkout speed while eliminating accidental or intentional losses.
- Grab and go. UltronAI can instantly capture which products customers have taken for faster, more accurate contactless checkout.
- In-store analytics. With better product identification, stores can improve sales and inventory management.
- Shelf monitoring. By embedding UltronAI into solutions for shelf monitoring, retailers can maximize profits by improving product availability and location.
“Given the rapid pace of change required to adapt and thrive in today’s market, retailers are working hard to incorporate transformative technologies into their businesses,” said Stefanos Damianakis, CEO of UltronAI. “With the availability of UltronAI’s product identification technology, retail innovators can reimagine how to build scalable and reliable solutions to combat shrinkage, improve customer experience, drive automation, and more.”
UltronAI is in the early stages of deployment with a leading global retailer and an inventive retail automation solution provider. During the testing phase, UltronAI successfully ingested the retailer’s 250,000+ product catalog in under 45 minutes, a stark contrast to the company’s former system which took days to weeks to enroll products and hit system capacity in the low thousands. In addition, as part of a real-world, in-store pilot deployment, UltronAI achieved accuracy comparable to a human against the store’s product catalog, despite variable lighting and positioning conditions.
