Mango Diseases Identification by a Deep Residual Network with Contrast Enhancement and Transfer Learning

Plant diseases have a great influence on the productivity and economics of agriculture. In the traditional method where manual labour and experiences play important roles, early plant disease detection and prevention for field crop are inefficient. This can bring negative effects in stopping the spread of diseases. Applications of image processing techniques and computer vision may provide a solution to these problems. Deep learning is known as one of the most powerful techniques which is able to address more complicated tasks compared to traditional machine learning thanks to the embedded complex layers. This paper presents an image-based diseases identification method using a deep neural network with contrast enhancement and transfer learning from the PlantVillage dataset. Since there exist various sizes of leaf, rescaling and centre alignment are performed to standardize images. Besides, contrast enhancement is used to improve quality of visualization which gives a well-prepared step for further processing. By using a collected mango disease dataset, the proposed model is trained to identify 3 common diseases from the healthy one. The proposed model is then compared with other pre-trained models. An accuracy of 88.46% is achieved for the proposed model, which is higher than the results obtained from other pre-trained models.