Fusarium head blight (FHB), a devastating disease of wheat and barley, can markedly reduce yield and grain quality. Current disease mitigation strategies including cultural practices, fungicide application, and planting resistant varieties rely on accurate and efficient phenotyping of FHB severity. Most projects that include field FHB screening utilize manual phenotyping methods, which do not provide adequate resolution for detecting small severity differences, is laborious, low throughput, and have rater bias and inter-rater variation. Increasing the throughput of FHB assessments in the field is necessary for continued improvement of varieties, efficacy of management practices, understanding disease development, and monitoring of FHB. We have deployed a sophisticated motorized phenotyping rover for high temporal and fidelity FHB detection in wheat and barley across resistance breeding and germplasm evaluation trials. This rover greatly increases the speed, accuracy, reliability, and automation of FHB severity assessment across experiments and scales. This past summer we used the rover to collect high-fidelity, field-based images over short time intervals (4 days) across hundreds of wheat and barley genotypes at two different field locations. After top and side images were acquired, machine learning algorithms are being used to identify spikes in each plot and develop highly accurate models for FHB detection and progression. Despite being at an early stage of development, these transformational, unique, and streamlined methods are markedly increasing the speed, accuracy, frequency of FHB phenotyping, and directly increasing the number of spikes evaluated per plot. The phenotyping pipeline and FHB assessment models generated from this work will be applicable to any researcher who could benefit from an easy to use, high throughput phenotyping rover for field FHB detection in wheat and barley. By combining high temporal and high fidelity field-based phenotyping, we envision eliminating previous phenotyping limitations and foresee greatly increasing the phenotyping capabilities of programs with respect to the amount of germplasm screened for FHB resistance, the number of spikes evaluated per plot, and the precision of measurements.