Plant Disease Detection Using Deep Convolutional Neural Network

In this research, we proposed a novel 14-layered deep convolutional neural network (14-DCNN) to detect plant leaf diseases using images. A new dataset was created various open datasets. Data augmentation techniques were used balance the individual class sizes of dataset. Three image used: basic manipulation (BIM), generative adversarial (DCGAN) and style transfer (NST). The consists 147,500 images 58 different healthy diseased classes one no-leaf class. DCNN model trained in multi-graphics processing units (MGPUs) environment for 1000 epochs. random search with coarse-to-fine searching technique select most suitable hyperparameter values improve training performance model. On 8850 test images, achieved 99.9655% overall classification accuracy, 99.7999% weighted average precision, 99.7966% recall, 99.7968% F1 score. Additionally, better than existing learning approaches.