New publication: Automated late blight lesion detection in the potato field

This new article associated to NordPlant identifies late blight lesions in field RGB images with the help of deep convolutional neural networks.

The article is published in Knowledge-Based Systems:

Junfeng Gao, Jesper Cairo Westergaard, Ea Hoegh Riis Sundmark, Merethe Bagge, Erland Liljeroth, and Erik Alexandersson. “Automatic late blight lesion recognition and severity quantification based on field imagery of diverse potato genotypes by deep learning

We retrieved RGB images of late blight infection from many different potato cultivars since disease lesions develop differently depending on genetic background together with Danish potato breeder Danespo, Jutland Denmark.

The size and ratios of the images was optimised for analysis and an automated majority voting algorithm between masks were employed. A good correlation between manual visual scores of and the number of lesions detected by deep learning vision at the canopy level.

It is a collaboration with University of Copenhagen and Danespo, and part of the pre-breeding PPP 6P2 as well as NordPlant.