From the start of farming as a profession, the weeds have been a problem and the problem still continue. The main factors are the competition for nutrients, water, and space and for disease and pesticides. Weed management is becoming popular in crop production with the arrival of high yield varieties and heavier doses of fertilizer application. One important way of improving fertilizer use efficiency is to control weeds in good time. We developed a web application which uses in-depth learning, to capture an input and to detect whether an image is normal or weed image. In this paper VGG16 classifier and RMSPROP optimizers are used for distinguishing a plant and a weed. The dataset is downloaded from Kaggle and Signal handling gathering of the Aarhus. Four plants, common wheat and sugar wheat, common weeds, cleavers with a totality of 942 pictures have been successfully classified. Finally, around 92% precision with a 5.5% rise in current classifications is projected by this work.