PhosopNet: An improved grain localization and classification by image augmentation

Rice is a staple food for around 3.5 billion people in eastern, southern and south-east Asia. Prior to being rice, the rice-grain (grain) previously husked and/or milled by milling machine. Relevantly, grain quality depends on its pureness of particular specie (without mixing between different species). For demand purity inspection an image, many researchers have proposed classification (sometimes with localization) methods based convolutional neural network (CNN). However, those papers are necessary large number labeling that was too expensive be manually collected. In this paper, image augmentation (rotation, brightness adjustment horizontal flipping) appiled generate more images from less data. From results, improves performance CNN bag-of-words model. future moving forward, recognition can easily done images.