Plant diseases are one of the significant threats to agricultural yield, causing economic loss and food insecurity. Manual inspection for detection is time-consuming, prone to errors, and inefficient for large monitoring areas. Although convolutional neural networks (CNNs) have proved helpful in automating the detection of plant diseases, the current methods still encounter issues corresponding to the generalization of CNN models in datasets of low quality and the prediction of disease progression. To overcome these limitations, this study presents an optimized Pytorch Custom CNN-based model for real-time and accurate plant disease classification. The significant contributions are focused on enhanced classification capabilities through refined deep learning architectures, incorporating CNNs within mobile architectures for better usability in real-life conditions, and an extensive performance evaluation compared to traditional methods. Through experimental results, the performance of the proposed model accuracy 96.13% which is better than that of existing approaches, significantly improving disease identification and management efficiency. The study enhances the application of artificial intelligence in agriculture, offers valuable information regarding sustainability and future research in plant disease monitoring, and will also help develop future precision agriculture.