Mango is considered as the King of Fruits and has various health benefits. Despite of the nutritional values that mango fruits possess; mango leaf disease is a common problem that effect productivity of mango crop. It is crucial to identify the plant disease for improving healthy and nutritional mango fruits. Visual inspection for identifying mango plant disease detection is a difficult task where deep learning can be used as a power tool for detecting and classifying mango plant healthy and unhealthy leaves. In this work, we proposed a novel method for identifying plant leaf detection using an ensemble deep learning model that uses gravitational search optimization algorithm. The preprocessing of the leaf images uses Cycle GAN augmentation to improve classifier performance. The proposed method evaluated on publicly available Plant Village datasets. We have chosen five deep learning models and compared their performance with the proposed ensemble convolution neural network. The proposed model achieved accuracy of 98.89% and F1-Score 94.5% for detection of mango crop leaf disease.