Plants are commonly used for treating many disorders since the golden ages. One of these plants is Rumex Nervosus that belongs to the family of Polygonaceae, which is traditionally used to treat various diseases in many countries such as, Yemen, Saudi Arabia and Ethiopia. The various types of plants that are existed in the Yemen make the recognizing of them a difficult task that requires knowledge and greet experience. The recognition of plants has a significant and crucial role in the classification of plants and differentiation between leaves. In this paper, an intelligent system is proposed to design a model that classify four types of plants (Rumex Nervosus, Agave, Green Grass, and Junipers) by using three pre- trained transfer learning models (AlexNet, GoogLeNet and VGG19) based on deep learning techniques. The process of plant recognition goes as follows: images of the four types of Yemeni plants are collected through a smartphone camera, where the total number of images of plants was 600 images for each class of 150 images. Then the images are pre- processed by resizing and center cropping images to fit the inputs of the proposed models. To improve the image recognition process, data augmentation has been performed, where the number of images is increased by creating different versions of content similar to images in order to obtain more training examples, as the number images reached 1700 plant images, and after that the images are forwarded to the proposed pre- trained transfer learning models (AlexNet, GoogLeNet and VGG19) that trained on ImageNet dataset by fine- tuning the proposed plants images. The proposed models automatically extract and classify these features, which help the network to recognize the type of plant effectively. The result of proposed models (AlexNet, GoogLeNet and VGG19) give accuracies (99.83%, 98.38% and 98.75%,) respectively. We found that AlexNet model gives the best result with accuracy 99.83%. Vividly, the proposed system was tested and compared to other works. the experimental findings show the effectiveness of the proposed method. The recognition of this type of plant accurately will help many populations to reliable recognition and fast treatment. In future work, we propose to improve methods for extracting features of plants using image segmentation techniques with machine learning techniques to recognize plant images with high efficiency and accuracy. We also suggest adding other models of deep learning techniques and making improvements in their structure.