Paper Title
A Convolutional Neural Network for the Identification of Diseases in Rice Crop

The pest and diseases are threatening to food security and cause significant yield losses in crops, especially rice. The improper and delayed diagnosis of disease resulted in unnecessary use of pesticides that increased the cost of production, environmental and health impacts. Therefore, timely and accurate diagnose through modern approaches is a dire need, for this purpose a study was conducted to develop a system for timely and accurate detection of rice diseases. The RGB high-quality images of rice diseases such as bacterial leaf blight, leaf smut, and brown spot diseases were collected from the various farm fields during 2019. The Rice diseases were classified by training the off the shelf standard convolutional neural network (CNNs). The pre-trained residual neural network (ResNet50) was taken from the ImageNet. The fine-tuning of these networks was performed on the collected data, 80% of data were used for training, and 20% for testing. The results showed that high accuracy was achieved from the ResNet50 which ranged from 75% to 91% in different batches. However, the Overall accuracy of ResNet50 was 83%. The learning curves showed that training and validation losses reduce overtime to a stability point with less difference to the final losses’ values, therefore, the model is considered to be well fitted and useful for the identification of rice diseases. Keywords - Disease detection, Rice, Convolutional Neural Network (CNN)