Recognising plant diseases using deep convolutional networks to analyse leaf photos is the main focus of this article. We train a deep convolutional neural network to detect crop illnesses using a large publicly available dataset made up of photos of healthy and sick plant leaves taken in controlled environments. The convolution method involves applying filters to an image in order to produce a feature map. The input picture or feature map is first passed through a linear filter with a bias added, then a nonlinear filter is applied. Following this, Max Pooling is used, which simplifies calculations for higher layers, removes the minimum value, and offers translational invariance. There are three methods in which our accuracy in determining whether tomato and potato leaves are healthy was proven: Early blight or the bacterial spot is contained inside. Our approach achieves an accuracy of 91% thanks to the given outcome.