Wheat Rust Disease Segmentation from Ground Imagery

Wheat rust disease is a major agricultural threat, causing significant yield losses globally. Precision agriculture and autonomous robot systems represent pivotal advancements in modern farming and disease detection, leveraging technology to enhance crop management, optimize resource use, and improve productivity. These systems rely heavily on accurate and comprehensive data to train effectively. However, current datasets often fall short of meeting the specific needs of these advanced agricultural technologies. This paper addresses this gap by introducing a primary contribution in the form of an open-source dataset focused on wheat rust disease segmentation for precision agriculture. Open-source datasets in literature are either general plant disease datasets or wheat disease classification datasets. Our proposed wheat rust segmentation dataset, Focussed Rust Disease Images (FRDI), provides close-up views of disease manifestations on wheat leaves, capturing the natural variability and complexity of diseases in field conditions, thereby meeting the requirements of precision agriculture systems. This facilitates detailed analysis and effective training of deep learning models designed for disease detection and segmentation tasks for real-world agricultural environments. Furthermore, in our work we present an improved wheat rust disease detection and segmentation technique which is based on the concept of co-salient feature extraction for disease identification within complex and cluttered field images. By integrating localized and generic contextual information, our proposed approach effectively identifies key features and intricate patterns associated with wheat rust disease.