Computer Vision News - January 2019

12 Computer Vision News Fruit Grading and Sorting with Deep Learning Every month, Computer Vision News reviews a successful project. Our main purpose is to show how diverse image processing techniques contribute to solving technical challenges and real world constraints. This month we review a precision agriculture project by RSIP Vision’s : Fruit Grading and Sorting with Deep Learning . This research is the result of a cooperation between RSIP Vision and Sunkist Growers - Research & Technical Services (RTS) , division of a leading not-for-profit marketing cooperative entirely owned by and operated for the California and Arizona citrus growers. Project One of the most interesting challenges in handling the supply chain of fruits and vegetables lays in the ability to sort and grade produce in the optimal way , given the whole range of possible product features and the requirements coming from the distribution channels to the packaging house. RSIP Vision has been involved in several projects within this application field. During the past three years, we have boosted precision agriculture work to drive it into the new Artificial Intelligence age. This specific project was conducted in partnership with Sunkist Growers RTS , division of a growers cooperative dedicated to run traditional growing practices while pioneering innovative solutions. Together, we launched a new approach to AI that enables to control result and at the same time solve any issue that might arise while performing this task. Sunkist RTS had developed a large sorting machine, with great field results. But they wanted to perfect them, by developing the machine of the next generation. Achieving the best possible grading results necessarily entails integrating new technologies like Deep Learning . A major requirement set by Sunkist RTS for such an AI system was that the Deep Learning networks would provide specific reports about specific features and events, rather that one unique result for the whole produce, which might be like a black box hiding the detailed results. Other challenges needed a solution, like time constraints : in agriculture, due to its nature of massive throughput needed by the clients, there is very little time to handle each produce. The fruit must be processed and graded in real time as they flow through the sorting/grading machine - many fruits per second .