Plant Disease Detection Using DeepNets and Ensemble Technique

Growing food security is a significant concern in the modern world. With the world's population expected to increase by two billion in the next three decades, there is a necessity to increase food production to support the growing population. In recent years, the increase in global food production has slowed, too slow to keep up with population growth. The factors directly affecting global food production are drought and plant diseases. Detection of these diseases through manual inspection is time taking and involves a factor of human error. In this paper, we focus on the problem of detecting plant diseases accurately at an early stage to increase food production. Machine learning and deep learning-based models have the potential to solve this issue by detecting plant diseases quickly and accurately. In this work, we first analyze the performance of pre-trained deep learning models on an expanded version of the standard PlantVillage dataset and then propose an ensemble of deep learning models. The proposed ensemble model outperforms all the existing deep learning models and achieves a maximum accuracy of 99.61%.