The PlantHealth project was selected in the framework of the call for Post-Docs launched jointly by Agropolis Fondation, the CEMEB and NUMEV labexes and the Convergences #DigitAg Institute in 2016-2017.
One of the main difficulties encountered in plant disease epidemiology is the lack of data. Moreover, the automatic recognition of plant diseases in crowdsourced image streams represents an important scientific challenge. The original approach proposed by the PlantHealth project to solve these problems consists in relying on deep learning and proactive learning solutions in order to set up a citizen science programme.
The selected post-doctoral fellow recently received the 2020 Best Publication Award in the journal Computers and Electronics in Agriculture for her paper "New Perspectives on Plant Disease Characterization based on Deep Learning".
The control of plant diseases is a major challenge to ensure global food security and sustainable agriculture. Several recent studies have proposed to improve existing procedures for early detection of plant diseases through modern automatic image recognition systems based on deep learning. In this article, we study these methods in detail, especially those based on convolutional neural networks. We first examine whether it is more relevant to fine-tune a pre-trained model on a plant identification task rather than a general object recognition task. In particular, we show, through visualization techniques, that the characteristics learned differ according to the approach adopted and that they do not necessarily focus on the part affected by the disease. Therefore, we introduce a more intuitive method that considers diseases independently of crops, and we show that it is more effective than the classic crop-disease pair approach, especially when dealing with disease involving crops that are not illustrated in the training database. This finding therefore encourages future research to rethink the current de facto paradigm of crop disease categorization.