Detecting plant diseases is a challenging and time-consuming task that requires expertise and laboratory conditions. Deep learning methods have been proposed as a solution to this problem, and their effectiveness in plant disease detection has become a popular research topic in recent years. This study aimed to investigate the performance of the U-Net architecture, which has been successful in medical image segmentation, in the segmentation of agricultural images. Sixty images, including angular leaf spot and bean rust diseases commonly found in bean plants, were used in the study. The images were segmented with U-Net, and then, in the first stage, only the images containing diseases were classified. In the second stage, classification was performed using raw images. Deep learning methods VGG16, AlexNet, MobileNet-v2, and DenseNet201 were used for the classifications. The results showed that the classification accuracy was higher for the segmented images than for the raw images. The highest accuracy rate, 100%, was achieved with DenseNet201 in the classification carried out by removing the segmented diseased regions. Using the U-Net architecture, which has demonstrated good performance with relatively few medical images, promising results were achieved in segmenting plant diseases. A software was developed to obtain only the images of the diseased areas by overlapping the original images with the segmented images. The proposed end-to-end system achieved higher classification accuracy by focusing deep learning architectures only on the desired regions. Finally, 100% classification accuracy was achieved with the DenseNet201 architecture using only segmented diseased images.