Livestock poisoning by several kinds of poisonous plants causes grievous economic losses to the livestock industry. Poisonous plants are also a fatal threat to humans, ingesting these plants can cause several side effects in the body because of their toxicity. Hence, it is essential to develop a rapid approach to recognize poisonous plants efficiently. This paper addresses a recognition approach for eighteen poisonous plants using poisonous plants leaf (PPL) dataset which has been generated using image augmentation techniques that contains 54000 training, 27000 validation, and 9000 testing images. Six different state-of-the-art deep learning models have been used in this study such as Xception, ResNet152V2, InceptionResNetV2, MobileNetV2, DenseNet201, and NASNetLarge for classifying leaf images of poisonous plants. Xception has shown more significant performance than other models, achieved 99.71% training and 99.37% testing accuracy. NASNetLarge and InceptionResNetV2 have achieved 96.89% and 95.18% test accuracy, respectively, and MobileNetV2 achieved the lowest test accuracy.