Early and precise detection of plant diseases is crucial for enhancing crop yield and minimizing agricultural losses. This paper evaluates the performance of deep learning-based Convolutional Neural Network (CNN) for automated plant disease detection. Four prominent CNN architectures - ResNet-50, AlexNet, GoogLeNet, and VGG - are trained and optimized using a labeled dataset of pepper bell plant images. The performance of the models is assessed based on standard evaluation metrics, including accuracy, precision, recall and F1-score. Among the evaluated architectures, ResNet-50 achieves the highest accuracy of 99.01%, demonstrating its effectiveness in differentiating between healthy and diseased plants. The results highlight the potential of deep learning models for real-time, automated agricultural diagnostics, advancing precision farming and sustainable agriculture.