This research aims to develop a mustard plant disease classification system using the Convolutional Neural Network (CNN) method integrated into a web-based platform. Classification is carried out on three classes, namely Spotted Mustard Leaves, Rotten Mustard Leaves, Healthy Mustard Leaves, with the addition of the Not Mustard Leaf class as a distractor class to test the robustness of the model against images that are not included in the main classification category. The dataset used consists of 800 images, 200 images each per class. The CNN model was built with a sequential architecture consisting of several convolutions, pooling, dropout, and dense layers, and using ReLU and SoftMax activation functions in the output layer. The training process is carried out up to 100 epochs, but with the use of Early Stopping callback, the training stops at the 60th epoch, with the best performance (best epoch) achieved at the 32nd epoch. Evaluation of the model on test data showed an accuracy of 93.75%, with high precision, recall, and F1-score values in each class. The model was then implemented into a web interface so that users could upload leaf images and obtain classification results automatically. The results of this study show that CNN is effective in detecting mustard leaf disease and has the potential to be applied as a digital image-based diagnostic tool in agriculture.