Weed Detection from the Duck’s Perspective Using Aqua-Drone in Rice Paddies

In research of weed detection, neural-network based algorithms are often used on both rice paddies and fields. In this research, we propose a method for detecting weed pixels using semantic segmentation from a front-facing camera mounted on a aqua drone, a duck’s perspective, in a paddy field as well as present a dataset of paddy field weed images for training neural network models. We have trained a semantic segmentation model and obtained the results of weed IoU: 0.383, mIoU: 0.676, Pixel Accuracy: 0.970.