TaiwanFood and restaurants2025 – ongoingHanlai Foods
AI food-waste recognition at buffet restaurants
A Taiwanese buffet chain weighed and photographed its food waste with AI and cut waste per diner by about 70%
Problem
Hanlai Foods, which runs 22 restaurant brands in Taiwan, produced about 1,101 tonnes of food waste in 2025. Taiwan's ban on feeding food waste to pigs from 2027 was expected to raise its disposal bill from roughly NT$400,000-500,000 a year to more than NT$10 million, and the kitchens did not know which dishes ended up in the bin.
Approach
Under a 2025 Ministry of Economic Affairs smart-dining demonstration program, Hanlai installed a custom machine at its Haigang buffet in Kaohsiung in October 2025: a camera above a scale identifies and weighs each tray staff throw away and ranks the most-wasted dishes. A local company built the hardware with three universities. Early recognition accuracy was only about 70%, so chefs log unrecognised items by hand, which also adds training data.
Results
Takeaway
Romaine-heart and bell-pepper garnishes topped the waste list, prompting cuts to plating, purchasing and late-service refills, with chef bonuses tied to savings. Diners' plate leftovers were excluded because mixed food was too hard to recognise.
Categories
- Industry
- Food and restaurants
- Business function
- Production and qualityLogistics and supply chain
- Innovation type
- Process innovation
- AI technology
- Computer vision
- Adoption
- Government support program
- Evidence
- Company's own claim
Sources
Source summaries
CNA reported that Hanlai's Haigang buffet cut food waste by about 1.6 tonnes in roughly three months and per-diner waste from 140 g to 42 g between September 2025 and February 2026. It notes early recognition accuracy of about 70%, with salmon once mistaken for grapefruit, and a plan to equip 18 buffet outlets by 2030 for about NT$10 million.
Business Next reported that a local hardware maker and three partner universities built the system. Hanlai chose a local build because the foreign product was hard to customise. Per-diner waste fell from 140 g to about 42 g. It describes bulky first-generation hardware, poor early recognition, chef bonuses linked to savings, and the decision not to track diners' plate leftovers.
Full details are in the original sources.
