The inspiration behind the project was seeing the excess food waste of certain dishes at UMass dining halls. Powered by our desire to improve sustainability in the community, we created this AI-based website to aid local chefs in reducing food waste and improving the quality of experience by optimizing their menus.
We created a small, manually labeled folder to pre-train data which was then utilized to autonomously annotate a larger food waste dataset. This was able to recognize key food groups which can be extrapolated for new images and checks to see the type and amount of food that got wasted. We designed a custom convolutional neural network from scratch that inputs the pre-trained data and improves itself through cost functions. This was then sent to a Raspberry Pi 4B that holds the data set and uses computer vision and an external web camera to automatically check new food images. Afterward, we implemented logic and web scraping from the dining halls to create specialized advice, robustly displayed on a custom-programmed website.
Log in or sign up for Devpost to join the conversation.