Agriculture is a vital source of livelihood and the backbone of the Indian economy. Diseases significantly impact plant production, but timely detection can improve health standards and economic growth. Crop protection is necessary to avoid diseases generated by bacteria, viruses, and fungi, and ensuring the upkeep of agricultural goods. Plant disease detection aims to monitor large crop fields and automatically detect disease symptoms as they appear on plant leaves. This paper utilizes Convolutional Neural Networks (CNNs) algorithm to detect 12 different plant diseases and their affected areas, providing proper fertilizers to prevent further damage from pathogenic viruses. It proposes a Custom CNN model with Laky ReLU function model for disease identification and classification. The process involves datasets gathering, training, segmentation, feature extraction, testing, and classification. The datasets of diseased and healthy leaves are trained using various CNN models along with different activation functions in order to compare the performance evaluation metrics to compare our proposed model. The model developed is established as a web-based application built using Pytorch Framework. The model, equipped on a controlled dataset of leaf images, achieved 98.7% training accuracy demonstrating the technical feasibility of CNNs in categorizing diseases, and paving the way for AI solutions for all the farmers.