This study describes the use of deep learning methods for segmentation and classification of jute leaves with respect to their age, in order to improve the speed and accuracy of disease detection. We utilize a publicly available dataset containing images of jute leaves categorized into three classes: Healthy, Cercospora Leaf Spot, and Golden Mosaic. Jute, being the most important cash crop in Bangladesh, is being threatened by diseases causing significant losses in quality and yield. The traditional way of disease detection, depending on manual inspection is often a laborious and flawed process. A number of state-of-the-art models, including VGG19, InceptionV3, Xception, ConvNeXtBase, and MobileNetV2, are scrutinized to determine the number of images that can be correctly classified also hyperparameter tuning was done on MobileNetV2 for better accuracy. MobileNetV2 model has the highest accuracy of 84% among those architectures. This is further retuned by hyperparameter tuning increasing the accuracy to 94% which is a big improvement. The reports demonstrate that the tuned MobileNetV2 model is faster and more accurate when it comes to precision, recall, and F1-score. The findings reveal that the optimization of hyperparameters is vital to the improvement of deep learning models for the classification of plant diseases. The study provides an innovative solution to the automated detection of jute leaf diseases; thus, the application of this technology may lead to the development of farming that is more efficient and sustainable