Monitoring plant status is one of the most important tasks in digital agro. It is important to determine the type of disease , as well as its location, that can be used for calculation of per-centage of plant damage. Our task is to fulfillthese two tasks through classification, segmentation and class activation map. We also present the updated Plant Pathology 2020 dataset withground truth masks of leaf diseases. For more details please see report.
We have captured high-quality, real-life RGB images of apple leafs with diseases from Plant Pathology 2020. Originally dataset contain 3642 images with class labels, there are 4 classes namely healthy, rust disease, scab disease, both diseases. We managed to get ground truth mask disease annotations for 1291 images from dataset (400 images for each of healthy, rust, scab and 91 images for both diseases class). We used only this 1291 images in our work.
https://www.youtube.com/watch?v=kpclmHUDHxo
