Enhanced Vision Transformer with Adaptive Segmentation for Precise Plant Diseases Detection

Early and accurate detection of plant diseases is of prime importance for agriculture and food production. Conventional methods in detecting plant diseases usually cannot guarantee speed and often tend to be inaccurate. In addition, recent AI models that have advanced the situation are still uncertain of where precisely to target the ailing parts of the plant, reducing their precision. Plant diseases threaten agricultural productivity, requiring early and accurate detection for effective management. Existing methods often lack precision in classification, localization, and segmentation. This project improves upon them using deep learning models—CNN for classification, Faster R-CNN for object detection, and U-Net for segmentation—on the PlantVillage dataset, which includes leaves, stems, and roots. this approach improves accuracy, disease localization, and segmentation compared to traditional methods. The results demonstrate the effectiveness of deep learning in automated plant disease diagnosis, contributing to smart agriculture and precision agriculture.