Timely identifying agricultural diseases is crucial for minimizing farmers’ economic losses and enhancing production efficiency. Bacterial spot, early blight, Septoria leaf spot, and late blight are common diseases affecting potatoes and tomatoes, two of the most extensively consumed staple crops worldwide. Early detection can mitigate production harm and optimize harvest yields. Avoiding plant diseases can enhance crop yield and reduce expenses. Enhanced sickness identification and management are attainable through contemporary technological advancements. This work proposes an efficient and cost-effective model inspired by deep learning methodologies for identifying agricultural diseases. This study employed convolutional neural network models to examine leaf photographs. The primary objective of this research is to acquire knowledge and build the capacity to diagnose illnesses in their early stages. The performance of the proposed model has been corroborated by trials performed on traditional pre-trained models. After the pre-processing of the input images, the designated area is isolated. The proposed architecture’s categorization performance was evaluated against various models, including Xception, MobileNet, and deep learning techniques such as ResNet50 and InceptionV3. Edges, colors, and textures are among the distinctive attributes that CNN can recognize in images. The experimental outcomes illustrate the suggested algorithm’s efficacy and competitive performance.