SmartCropCNNLite: Real-Life Tomato Crop Disease Detection Using Deep Learning and Smart Devices

Abstract

Today agriculture industry is going through a fundamental transformation to adopt “High-Tech Farming” and “Precision Agriculture” leveraging sophisticated technologies such as robotics, IoT sensors, Computer Vision, GPS, and aerial images to improve crop yield. However, it is projected that the world population will reach 9.9 billion by 2050 from the current 8.1 billion population in the year 2024. Currently, nearly 800 million people suffer from hunger worldwide and 8% of the overall world population is expected to be undernourished by 2030. The agriculture industry is continuously challenged by climate changes and the draining of natural resources. Crop yield security is threatened by various diseases, which are caused by fungal, viral, and bacterial organisms. This research focuses on the detection of six types of adverse diseases that are commonly impacting Tomato crops using Computer Vision and Deep Learning classification methods. The research takes advantage of the proven benchmark performance of simple and lighter Convolution Neural Network (CNN) models with Transfer Learning. To suggest appropriate treatment, this research further predicts the degree of disease spread using Machine Learning and Deep Learning techniques. The trained CNN model deployed onto Android smartphones can detect tomato crop diseases in the actual farm (plantation) field with acceptable performance. The solution implemented as a part of this research can allow small and mid-scale farmers to detect the plant diseases along with disease severity to control disease spread. This research has the potential to enhance tomato crop yield per hectare, directly influencing global GDP and making a valuable contribution to the overall agricultural economy.
Keywords: AI in agriculture, Computer Vision, Deep Learning and CNN, Real-time Tomato Crop Disease Identification, Disease Severity Detection, Android Mobile App