A Machine Learning Approach to Classification of Okra

Multiple machine learning techniques have been used for image classification purposes in agriculture. They can be applied to either roots, leaves or plants' detection and classification in order to assist farmers tasks. This paper proposes an image-based Okra classification for Japanese farmers in Okinawa. Thus, we implement Deep Learning to classify okras into categories pre-established by the Japan Agricultural Cooperatives. The classification features of okras in this study are their length and shape, and they classified into two: Class A and B. A set of pre-processing layers such as background noise cancellation, gray scaling and enhancement, image resizing and reconstruction are utilized to provide a higher detection rate. Moreover, a Convolutional Neural Network (CNN) is implemented to detect the patterns and predict the outputs.