Performance of a Working System Consisting of Garlic Trimmer and Sorter

Abstract

In this study, we developed a garlic sorter and experimentally evaluated the performance of a working system consisting of a garlic trimmer and a sorter. The garlic sorter detects trimmed roots and sorts them into three containers corresponding to three classes, namely, good, bad, and scratched using a deep learning model. If a trimmed garlic is sorted into the bad class, it will be trimmed and detected again until it is sorted into the good class. The results of testing the working system showed that the detection and sorting by the sorter were faster than those by an expert. The final pass rate of the good class was high irrespective of the experience of the operator.