The integration of machine learning techniques into modern agriculture has accelerated significantly, enabling more efficient and accurate crop monitoring practices. Deep learning architectures, particularly pre-trained algorithms such as AlexNet, ResNet, VGGNet, and GoogleNet, have demonstrated strong capabilities in precision agriculture applications, including plant disease detection, nutrient assessment, and growth analysis. These applications rely on heterogeneous data sources, ranging from smartphone imagery and UAV-captured images to satellite and sensor data, making analysis both data intensive and computationally challenging. This study introduces a transfer learning based framework that leverages multiple pre-trained models alongside a custom CNN to classify wheat diseases. A dataset comprising over 5,000 images, representing four disease categories and one healthy class, is utilized for training and evaluation. Model performance is assessed using standard metrics such as accuracy, precision, recall, and F1-score. Experimental findings indicate that the ResNet50 based approach achieves superior performance, reaching an accuracy of 93%, highlighting its effectiveness for agricultural disease diagnosis.Proposes a unified transfer learning framework combining pre-trained CNNs and a custom model for wheat disease classification.Utilizes a diverse dataset of 5,000+ images covering multiple disease classes and healthy crops.Demonstrates superior performance of ResNet50, achieving 93% accuracy across evaluation metrics.