| The global agricultural landscape faces a critical challenge in combating Fusarium Head Blight (FHB), a devastating wheat disease that threatens food security and farmer livelihoods worldwide. Despite technological progress, traditional disease detection methods, often relying on labour-intensive and error-prone manual inspection, have remained largely unchanged for centuries. In this research paper, we propose a deep learning-based convolutional neural network (CNN) for automated Fusarium Head Blight detection in wheat crops that leverages data augmentation and class weighting strategies to mitigate challenges associated with dataset imbalance. By leveraging advanced data augmentation and intelligent class weighting strategies, our proposed model transcends traditional limitations, achieving an impressive 96.36% accuracy on training data and 94.12% on validation data, with precision scores of 97.96% and 96.27%, respectively. These results demonstrate the model's robust performance and strong generalization capabilities and highlight its potential to enhance precision agriculture in an era of increasing climate uncertainty. As global food production becomes increasingly vulnerable to environmental challenges, this research represents a step towards empowering farmers with AI-driven tools that can quickly, accurately, and cost-effectively identify crop diseases, ultimately contributing to more resilient and sustainable agricultural practices. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.