Deep learning-based multi-spectral identification of grey mould

Early detection of economically important plant diseases, such as grey mould caused by Botrytis cinerea, is major importance for the timely application disease management strategies and reduction impacts on crop production environment. In this study, artificial inoculation leaves cucumber plants with B. cinerea under controlled environment was performed. Multi-spectral imaging used to capture fungal spectrum response at 460, 540, 640, 700, 775 875 nm, laveraging both RGB Near Infrared (NIR) channels. Two annotated image datasets were created from collected multi-spectral images named Botrytis-detection Botrytis-classification. Several deep learning-based classification object experiments conducted datasets. Classification results indicated that learning models can separate two classes accuracy 0.93 (F1-score 0.89), while achieved a mean average precision (mAP50) 0.88, paving way future early cinerea.