Restaurant Chain Achieves High Accuracy on Food Measurement With Computer Vision Solution
In this case study
In response, World Wide Technology (WWT) initiated an innovative computer vision proof of concept (POC) for the National Restaurant Chain employing the Plainsight vision AI platform, Google Cloud and a unique funding model provided by longtime WWT partner Intel.
The POC solution, deployed quickly, trained, and validated with only a small amount of POC data, demonstrated an ability to measure food quantity with an impressive 95% initial accuracy, providing real-time data and the potential of actionable insight to ensure customers' favorite food items will be available. Based on this first phase of success, WWT stands ready to expand the solution to other use cases beyond food measurement.
This privately owned National Restaurant Chain has grown to thousands of locations on its value proposition of fast, fresh, high-quality regional cuisine refined for American tastes. In-store guests proceed to the service counter to choose their entrees and side options. Behind the service counter, restaurant associates serve up each order in the required portions from serving stations on the steam table onto individual plates for in-store dining or takeout.
In a busy restaurant with multiple culinary choices, keeping the serving stations replenished with fresh-made items is critical. But, it can pose a food management challenge in estimating how many servings remain and when to start cooking replacement entrees without cooking too much food, which can contribute to food waste if not managed well. In the past, the decision of when to cook more food was the responsibility of the associates themselves, which required guesswork and was susceptible to human error, potentially leading to shortages of popular items that would force guests to either wait for replacement servings or choose something else.
To improve upon food management, the National Restaurant Chain initially installed weight sensors to calculate the remaining amount of food to better identify when serving stations might need replenishing, but that system proved to be problematic: Mechanisms became sticky from spilled food, diminishing their accuracy, and serving stations moved throughout the day. The varying serving station sizes, combined with the relative weights of different dishes, contributed to miscalculations as well.
Managing the balancing act of fast and fresh and plentiful is critical to the National Restaurant Chain's promise – but minimizing the waste from end-of-day food overages is important, too. All these factors underscore the importance of accurate food measurement, eventually leading to predictive orders for chefs to prepare more food.
Beyond the in-store experience, that same need for accuracy extends to other service options in the post-pandemic age: In-store dining, drive-thru, curbside pickup, delivery and online ordering, to name a few.
The National Restaurant Chain approached WWT for a technology solution to improve the accuracy of its food measurement practices. WWT worked with technology partners Plainsight, Google Cloud and Intel to devise a proof of concept (POC) for an end-to-end computer vision solution incorporating artificial intelligence and other emerging technologies to provide real-time food measurement data with eventual predictive capabilities.
In a food service environment, computer vision is instrumental in automating essential tasks and performing moment-to-moment analysis to manage operations – monitoring, tracking, alerts and more – with high degrees of precision. Plainsight enables customers in wide-ranging industries to develop, scale, and manage computer vision solutions that maximize the value derived from still images, video feeds and other visual sources, employing intelligent, intuitive systems.
Plainsight's computer vision solutions and end-to-end vision AI platform streamlined mechanized and manual operations alike, harnessing the power of AI. Given these capabilities of computer vision, WWT believed that Plainsight's technology offered the ideal solution for the specific challenge facing the National Restaurant Chain.
For the POC solution, WWT and Plainsight employed the newest technologies in AI, image capture, and infrastructure to deliver automated analytics for greater predictive capabilities.
A camera was installed over a serving station on the steam table, placed unobtrusively beneath the sneeze guard and trained on a food container to capture real-time video feeds. In addition, WWT deployed a vision AI model to monitor and calculate container fill levels. This regression model used the existing serving station-weighing system in initial training to confirm or adjust the weight measurements.
Once accurate results were achieved by the model, the visual AI-powered technology could be used alone to achieve highly accurate results. In addition, Google Cloud resources were leveraged to facilitate Plainsight's fusion of model training and sensor-based weight data to correctly identify the remaining quantity of each specific menu item.
In our consultative role as digital transformation specialists, WWT brought together the ecosystem resources to strategize and execute this POC success, melding physical spaces and digital data for real-world quality improvements. WWT, a frequent participant in Intel's history of supporting emerging technology innovations, approached Intel with this National Restaurant Chain POC opportunity.
By providing critical POC funding as part of its continuing technology alliance with WWT, Intel acknowledged the strong potential for this future-forward solution. In doing so, Intel supports our joint focus on new edge AI and analytics solutions that offer high-performance, cost-effective business transformation based on Intel® technology platforms wherever precision and predictability are key.
WWT was able to validate the AI-powered computer vision solution with a small amount of POC data, with bottom-line benefits that included:
- 95% initial accuracy of food measurement, minimizing the potential for human error, eliminating reliance on weight sensors and ensuring food items could be properly replenished before shortfalls occurred.
- Potential to increase sales – The long-term vision is to monitor all food levels so that favorite entrees are kept in consistent, reliable supply – an achievement of revenue growth that could be achieved with significant implications if the POC were rolled out to all of their thousands of store locations.
- Speed of implementation and validation – Thanks to the computer vision automation built into the vision AI platform, and to Plainsight's close alliance with Google Cloud as part of its overall agnostic offering of cloud service providers, WWT was able move quickly to implement the solution.
- Potential to enhance customer satisfaction as intelligent food management technology helps ensure a consistent, timely supply of the fresh, high-quality entrees that are foundational to the National Restaurant Chain's brand promise.
This powerful alliance – the digital transformation expertise of WWT, the AI-powered computer vision platform and solutions from Plainsight, the resources of Google Cloud, and the edge technology leadership and funding support of Intel – have been harnessed for a significant proof of concept that has the power to extend predictive analytics far beyond food management.