AI-Based Quality Control System at the Pressing Stages of the Champagne Production

Deep learning (DL) is a hot trend for object detection and segmentation, thanks to Deep Neural Networks (DNNs). Image recognition is a powerful tool for precision viticulture, having strong potential in yield estimation and automatic quality estimation of the grapes. However, developing the models is one part of the problem; deploying them in the field, at the edge of the network, is another problem that comes with its own constraints. This paper studies the use of embedded devices to run DNN algorithms for real-time grape segmentation at the wine press. The results show that it is possible to use edge devices while respecting a real-time context with little detection quality losses.