Research Projects

Plant diseases severely damage crops, accounting for over 50% of yield losses. To prevent these diseases, early detection and preventative measures are crucial. Nucleic acid-based and serological techniques are used for routine detection, while point-of-care diagnostic technologies are needed for quick, outside-the-labeled diagnosis. Advances in aerial imaging and noninvasive/non-image sensors have led to automated plant disease detection. These technologies collect data from various angles to study leaf phenotypes and identify plant diseases. Machine learning approaches, particularly deep learning models, have produced the best results for visual recognition concerns. Combining multidisciplinary approaches and increased availability of spatial, spectral, and satellite photographs provides a cost-effective approach for crop-disease classification techniques. However, large datasets are required to train deep learning models, which can be expensive and challenging to acquire. This study aims to develop and integrate an end-to-end deep learning approach as a lightweight mobile application for automatic recognition and monitoring plant diseases using transfer learning. The application will use technologies to miniaturize deep learning models for mobile phones and enable farmers to access an expert knowledge base to learn about the disease, its effect, and intervention mechanisms. The study focuses on the chilli leaf curl virus and groundnut early leaf spot diseases, which represent significant economic losses to farmers. Early detection using an automated system and its severity with suggested intervention mechanisms could be advantageous for increasing production and productivity.