One year of OPTIMA!
Imaging of diseased and control plants for the creation of the intended dataset
Wageningen University & Research performed greenhouse and field experiments and built a large dataset by imaging diseased and control plants using a high resolution RGB camera (10Mp), a six band multispectral camera (AirPhen) and a 150 band imaging spectrograph (IMEC SnapScan).
Apple trees with scab were imaged in a trail field close to Wageningen in the Netherlands. Greenhouse trails of grapes with downey mildew were imaged at the University of Torino. Field trails of grapes were imaged at commercial vineyards of Terre Da Vino in Italy. Greenhouse trails of carrots with Alternaria leaf blight were imaged at Agrocampus Ouest in France.
Analysis of the data gives an impression on the most discriminating wavelengths for the different diseases. Furthermore a deep learning pipeline was build, based on Yolov3 (You only learn once) with promising results on detection of downey mildew on RGB images of grape leaves imaged in the field.