Today, the automated classification of plant species is done by a variety of deep machine-learning methods. In order to improve the performance of feature extraction and classification for the deep learning models in plant classification, a model for classification of plants based on spatial domain enhanced segmentation with modified VGG-19 is proposed. The model was improved to make it more suitable for the recognition of plants from its leaves. Segmented leaves superimposed with original can greatly extract useful features only and reduce input parameters for the VGG 19 network. The experimental results suggest that the proposed method is optimal for classifying plants using segmented, superimposed as well as normal leaf images. The result shows state-of-the-art performance on the two processed data sets while remaining competitive in accuracy and computing time.