Improved Detection of Fusarium Head Blight in Wheat Ears through YOLACT Instance Segmentation

The Fusarium fungus, specifically Fusarium Graminearum, is the primary cause of Fusarium Head Blight (FHB), poses a significant menace to wheat, barley, and maize cultivators. Beyond manifesting symptoms like head discoloration and bleaching, FHB cunningly generates mycotoxins, detrimentally impacting grain quality, market appeal, and sustainability. Despite farmers' commendable use of fungicides, eradicating Fusarium Head Blight (FHB) remains a formidable challenge. The goal of this project is to perform a variety of research and develop novel tactics to aid in the ongoing fight against FHB. The goal is to provide a technology-driven solution for accurate yield estimation and selective harvesting, with a focus on the latter growth phases. To achieve accurate wheat ear segmentation from the background, the YOLACT model integrates modern tools such as RGB imaging captured in diverse environments. Significantly, the utilization of deep learning, a key aspect of our approach, involves color image processing to enhance FHB detection accuracy in the field. Highlighting that our proposed approach achieved an impressive accuracy of 98.81%. The proposed approach aligns with a forward-thinking perspective valuing innovation to mitigate the detrimental effects of this ubiquitous fungus on agriculture and the economy.