ABSTRACT
Crop losses caused by the spread of various diseases pose a serious danger to the livelihoods and food security of thousands of rural communities around the world. Crop disease detection is a task that a variety of approaches can perform. While many conventional approaches are quite popular and effective, image processing and now deep learning-based approaches are proving to be quite efficient and cost-effective. In this work, we focus on the major crops of India. We consider the major crops from all seasons and categories and discuss the various diseases affecting these crops. We review the deep learning approaches from 2012, which is when the current phase of deep learning started, and provide details of all the architectures and models developed over the years. These models and architectures are categorized based on their base neural network, application, and size. We also provide the details of open source and licensed datasets for crop disease detection for these crops. We compare the efficiency of all these architectures and models, and the review will help the research community decide the appropriate model to carry forward research in this field.
