Water Sprouts Detection of Cacao Tree Using Mask Region-based Convolutional Neural Network

Water sprouts are one of the causes of cherelle wilt in cacao plants, thus water sprouts must be pruned regularly. On the other hand, the number of cacao workers is increasingly reduced because young people prefer to work in urban areas. Therefore, an automation system to prune water sprouts on cacao plants is needed. Machine vision technology plays an important role in detecting water sprouts in the automatic pruning system in the plantation area. In this paper, the Mask R-CNN (Region-based Convolutional Neural Network) method is used to detect and segment water sprouts from images taken at a cacao plantation. The obtained data consist of 150 images which are for a training dataset of 120 images and a testing dataset of 30 images. To determine the system performance, the threshold parameter in the detection step was tested from 0.1 to 0.9. The best results are obtained at a threshold of 0.6 with an F1score of 0.907. This result shows that the Mask R-CNN method is able to detect water sprout properly.