FHB (Fusarium Head Blight) is a serious fungal disease that affects wheat and barley and lowers productivity because the affected grains are tiny, shrunken, of poor mass grain quality, and contaminated with mycotoxins that are harmful to animal and human nutrition. It requires precise and innovative methods for detecting and predicting disease symptoms at the earliest stages of pathogenicity. Traditional disease assessment of host plant genotypes is laborious, subjective (based on visual inspection), time-consuming, and costly, making it a limiting factor in plant breeding research. This study proposes a novel digital method for field-based FHB evaluation that makes use of a custom-built imaging device capable of taking close-up images of wheat heads from various angles. Ground-truth data were generated using polygon annotations of healthy and diseased spikelet in Roboflow. For model development, two deep learning pipelines were integrated: Attention U-Net for pixel-level segmentation and YOLOv11 for spike detection and disease classification. Together, they form a hybrid inference pipeline that quantifies disease severity per spike and row, generating masks, bounding boxes, and severity tables automatically. Technology automatically detects healthy and infected spikes by collecting high-resolution images from wheat plots and employing a complex predictive pipeline. Rather than relying on manual observation, the platform employs image-based learning to recognize disease symptoms and estimate infection severity. Results are visualized in an interactive web dashboard built using Flask, which allows users to upload field images, view colored severity maps, and download CSV or PDF reports within seconds. The process converts the raw images into quantitative summaries and color-coded maps indicating the percentage of FHB present in the wheat head. A web-based interface can help breeders, researchers, and producers use their tools more efficiently. Users can compare severity patterns across rows or plots, receive condensed reports, and evaluate analysis results immediately by uploading an image. The platform reduces the time between data collection and field decision making by providing feedback in almost real-time. This research demonstrates how cognitive computing and automated imagery can be used together to monitor crop health. It enables breeding operations to choose disease-resistant types more quickly by providing a dependable and scalable alternative to manual scoring. Future research can apply the method to other crop diseases, which would be a step in the direction of more intelligent, data-driven agricultural management.Keywords: FHB; Deep Learning; Attention U-Net; YOLOv11; 360-PICS; Disease Detection and Severity Estimation; Precision Agriculture