Cassava is the third largest source of carbohydrates for human food in theworld but is vulnerable to virus diseases, which threaten to destabilize foodsecurity in sub-Saharan Africa. Novel methods of cassava disease detection areneeded to support improved control which will prevent this crisis. Imagerecognition offers both a cost effective and scalable technology for diseasedetection. New transfer learning methods offer an avenue for this technology tobe easily deployed on mobile devices. Using a dataset of cassava disease imagestaken in the field in Tanzania, we applied transfer learning to train a deepconvolutional neural network to identify three diseases and two types of pestdamage (or lack thereof). The best trained model accuracies were 98% for brownleaf spot (BLS), 96% for red mite damage (RMD), 95% for green mite damage(GMD), 98% for cassava brown streak disease (CBSD), and 96% for cassava mosaicdisease (CMD). The best model achieved an overall accuracy of 93% for data notused in the training process. Our results show that the transfer learningapproach for image recognition of field images offers a fast, affordable, andeasily deployable strategy for digital plant disease detection.
展开▼
