Plants that grow are exposed to a variety of diseases. One of agriculture’s most difficult issues is the early detection of plant diseases. If infections are not diagnosed early on, they might negatively influence overall output, resulting in a loss of farmer earnings. To overcome this issue, some researchers developed sophisticated systems using Machine Learning (ML) and Deep Learning (DL) methods. However, millions of training parameters are needed for many of these algorithms, and their classification accuracy is low. With the help of a convolutional autoencoder (CAE) and a deep convolutional neural network (DCNN), this research offers a novel hybrid model for the autonomous detection of plant diseases. No state-of-the-art system has, as far as we know, shown how to create a hybrid system that employs CNN and CAE independently for the detection of plant diseases. In this work, the suggested hybrid model is used to use leaf pictures to diagnose Bacterial Spot disease in peach plants. The model achieved a classification accuracy of 98.7%, outperforming conventional CNN architectures. Additionally, the precision for major disease classes consistently exceeded 94.49%. The model’s computational time was optimized, reducing it by 140.67 ms compared to baseline models due to the efficiency of feature reduction by the CAE.