AWS Public Sector Blog

Zero Waste Zero Hunger: Using data and AI to reduce food waste in South Korea

Each year, the world wastes a staggering one-third of the food intended for human consumption. According to the Food and Agricultural Organization of the United Nations, South Korea wastes more than 130 kilograms of food per person annually, compared to 95 kilograms per person in Europe and 115 kilograms per person in North America. What drives this stark difference?

“The high amount of food waste stems partly from the fact that in Korean food culture, side dishes are an essential part of a Korean meal,” said Seunghee Lee, team leader of Busan IT Industry Promotion Agency. “However, in South Korean cafeterias, there is often no choice regarding the types or amounts of side dishes, resulting in many going uneaten.” Workers in Korean cafeterias, such as those in schools and businesses, have long been concerned with food-waste disposal costs. However, these cafeterias lacked a data-based system to better forecast diners’ consumption habits and demand, and the waste cycle continued.

The Busan Cloud Innovation Center (CIC) powered by Amazon Web Services (AWS) saw an opportunity to address this problem. The AWS CIC Program provides an opportunity for nonprofits, education institutions, and government agencies to collaborate with other public sector organizations on their most pressing challenges, test new ideas with Amazon’s innovation process, and access the technology expertise of AWS. The Busan CIC teamed up with an international humanitarian organization, the Busan IT Industry Promotion Agency, and technology startup Nuvilab. Using Amazon’s “Working Backwards” approach, the Busan CIC delivered two innovation workshops and conducted more than 40 end-user interviews to fully understand the food waste problem and develop a creative solution with technology powered by the cloud.

Leveraging artificial intelligence (AI), machine learning (ML), and data lake solutions on AWS, the team developed Zero Waste Zero Hunger (ZWZH), a program that uses AIML technology to provide data about food consumption.

Calibrating portion sizes with AI to reduce waste

At the heart of the ZWZH program is an AI-based 3D food scanner, built by Nuvilab, which analyzes food consumption and provides insights to both cafeteria managers and diners. When diners select plates from a cafeteria, they scan their plates with the Nuvilab scanner in the cafeteria before and after their meal. The Nuvilab scanner uses AI and ML to analyze the leftover food, identifying food waste by type and cost, gathering quantitative data that helps cafeteria managers optimize food inventory, quantity, and menu options. Additionally, the Nuvilab scanner is connected to a ZWZH dashboard in the cafeteria, which displays the volume of greenhouse gas reduction in real time whenever it detects a zero-waste plate. This gives diners insight into how much they’ve contributed to carbon emission reduction, which can lead to positive behavior changes.

“Waste is generated because kitchen managers cannot accurately identify customers’ preferences for each menu,” says Logan Kim, CEO of Nuvilab. “With our scanner analyzing food waste, managers can use these insights to design different menus or adjust recipes, improving food quality while minimizing wastage.”

The Busan CIC launched ZWZH in October 2020, testing it in cafeterias at six K12 schools and Korea Agro-Fisheries & Food Trade Corporation. Soon after, it deployed the program in two universities for pilot testing, completing the build phase in August 2021. The solution underwent an iterative refinement and validation process, with complete migration to the AWS Cloud in June 2022. Through adaptive ML model training ‘on the job,’ the 3D scanner now recognizes the type and quantity of food with 95% accuracy and serves 10 corporations and more than 40 national K–12 institutions.

Leveraging AWS to build and scale

The ZWZH solution relies heavily on AWS Cloud technology, but building the solution began with people, not tech. The Busan CIC team began by identifying the right sources from which to collect the most valuable data. For example, there were different types of food consumption to consider—not only in cafeterias, but also in retail restaurants, homes, and hotels. Lee’s team tackled this challenge by conducting interviews and surveys with samples of customers, restaurant owners, and cafeteria operators. They also analyzed each use case’s market demand and profit and loss (P&L).

Once the team identified its target locations for the initial launch, AWS helped the Busan team reduce the build-phase timeline by using AWS’s high-performance graphics processing unit (GPU) in Amazon Elastic Compute Cloud (Amazon EC2). Amazon Simple Storage Service (Amazon S3) stores the scanned food images, and the images are sent to GPU servers, and then AI models uploaded into the GPU analyze the food image to recognize the amount and type of each leftover food. The analysis results are stored in the Amazon Relational Database Service (Amazon RDS) and Amazon CloudFront sends the accumulated data to provide insights to cafeteria operators, restaurant owners, and other end users.

Reducing waste to make an environmental impact

The potential environmental impact of the ZWZH is significant. In South Korea, cafeterias implementing ZWZH reduced overall food waste by an average of 30%, while on-plate food waste was reduced by up to 42%. Estimates show that if every cafeteria implemented the ZWZH 3D scanning solution for 300 days with 500 consumers and prevented 300g of leftovers per plate from being created, cafeterias could reduce 15.3 tons of CO2—the equivalent of planting 1,848 pine trees per year.

Thus far, K–12 schools have seen the most significant benefit of the solution. Students can see how much food waste they generate, making them more conscious of waste reduction. Teachers can also utilize ZWZH as a visual aid to educate students about the environment. Students have reported being more conscious about their contribution to environmental protection and trying not to leave food on their plates after using the 3D scanner in the cafeteria.

But it’s not just schools realizing the waste-saving benefits of ZWZH. The Korean Ministry of National Defense is currently one of Nuvilab’s largest customers. With over 2,500 soldiers and staff served daily, the Ministry kitchen patrol workers used to serve each person around 110 grams of rice, based solely on personal experience before using the ZWZH solution. After implementing the solution in four battalions across Korea, the Ministry reduced the portions to an optimal 98 grams, reducing waste and saving costs, allowing them to spend more on other ingredients and fostering the health of their soldiers.

Evolving for the future

“Food waste is a global issue that has become more serious as food prices soar yearly, and organizations are receiving regulatory pressure to reduce food waste,” notes Lee. “ZWZH’s AI-based 3D food analysis technology powered by AWS, can be replicated in other regions and adapted to different food cultures to help reduce food waste and critically impact the environment in meaningful ways.”

Visit the Busan CIC main page to learn more about how cloud technology is supporting innovative solutions to improve lives. Learn more about how Nuvilab is using data and technology to help organizations reduce food waste.

Learn more about how AWS helps organizations positively impact the environment and reduce carbon footprints.

Read more about solutions from AWS Cloud Innovation Centers:

Subscribe to the AWS Public Sector Blog newsletter to get the latest in AWS tools, solutions, and innovations from the public sector delivered to your inbox, or contact us.

Please take a few minutes to share insights regarding your experience with the AWS Public Sector Blog in this survey, and we’ll use feedback from the survey to create more content aligned with the preferences of our readers.