Commercial kitchens throw away a great deal of food, and almost none of them know what. Waste is recorded, if at all, as a single weight at the end of a shift, which is far too coarse to change a menu.
This project built a fully automated food waste management system. A camera device captures the bin the instant something is discarded, the image is preprocessed in real time, and machine learning classifies and segments what was thrown away, estimating its weight from the difference against the previous state.
The output is the thing kitchens can act on. Not “42 kg of waste this week” but which dishes, which components, and on which days. Run by ZHAW’s Institute of Embedded Systems with the Centre for AI and the Institute of Computer Science, with Kitro SA.