Study of Segmentation Networks in the Detection of Ringspot Virus

Brazil is one of the main centers of papaya cultivation in the world. Papaya ringspot, caused by Papaya ringspot virus, is one of the main phytosanitary problems of this fruit tree, being responsible due to the itinerant nature of culture. Currently, in producing regions, its control is carried out through visual identification of symptoms by a rural worker who covers the entire area, with subsequent elimination of plants with symptoms of ring spot, a practice called roguing. The work presented here aims to carry out a study using image capture using ARP to generate a dataset of plantations containing the disease; from the images collected, label healthy, diseased plants and soil, so that we can train this dataset on segmentation networks (UNet, PSPNet, and LinkNet) combined with VGG16 and thus analyze the feasibility of using these networks to detect the disease in plantations. With the numerical and visual results presented here, detecting the disease with the networks trained here was satisfactory.