Creation of Farmers' Awareness on Fall Armyworms Pest Detection at Early Stage in Rwanda using Deep Learning

摘要

In recent years, research has proven that cutting-edge technology is being used to revolutionize the agriculture sector more effectively and efficiently in both advanced and developing countries. A great number of researches have been conducted, very little research, however, has addressed the use of machine learning in Rwanda to improve farmer's knowledge and skills to be able to tackle agricultural issues related to crops pests. In this regards, we introduced the creation of extension material in the local language "Kinyarwanda" to enable smallholder farmers to find out the right pesticide to use after identifying the pest, how to fight the infection and how to prevent it from spreading. In order investigate the aforementioned method and proposed technology; we did binary classification where we collected unhealthy plant leaves images from the fields to be tested. We again obtained healthy plant leaves images without damage from a very well-known research and development unit called Plant Village where we collected our data. The findings of the current study are based on the trained model with Keras. The result of the classification model provided a satisfying accuracy 0f 99.5% and that model was deployed in an Android Mobile App to be used for both pest detection and creation of farmers' awareness on suitable crop pests' detection and diagnosis in a timely manner. Our findings contribute to the use of machine learning in agriculture literature and have important research in deep learning and practical implications.