Mobile Platform Implementation of Lightweight Neural Network Model for Plant Disease Detection and Recognition

Due to the increasing world population, agriculture sectors from around the globe are challenged to increase their yield per year. However, harvests suffer from defects due to plant diseases. The current methods for mitigate spreading plant diseases are entirely dependent on the detection and recognition of such. Detection and recognition systems for plant diseases often require huge database for reference and/or computationally expensive systems. In this paper, we present a computationally light neural network model for detection and recognition of plant diseases and implement it to a mobile platform. Here, a two-step training process is used: pre-training on ImageNet data set of wide variety of objects and retrained on data set of specific plant diseases. The model achieved a test accuracy of 89.0 %.