English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Deep learning techniques have been successfully deployed for automating plantstress identification and quantification. In recent years, there is a growingpush towards training models that are interpretable -i.e. that justify theirclassification decisions by visually highlighting image features that werecrucial for classification decisions. The expectation is that trained networkmodels utilize image features that mimic visual cues used by plantpathologists. In this work, we compare some of the most popularinterpretability methods: Saliency Maps, SmoothGrad, Guided Backpropogation,Deep Taylor Decomposition, Integrated Gradients, Layer-wise RelevancePropagation and Gradient times Input, for interpreting the deep learning model.We train a DenseNet-121 network for the classification of eight differentsoybean stresses (biotic and abiotic). Using a dataset consisting of 16,573 RGBimages of healthy and stressed soybean leaflets captured under controlledconditions, we obtained an overall classification accuracy of 95.05 \%. For adiverse subset of the test data, we compared the important features with thoseidentified by a human expert. We observed that most interpretability methodsidentify the infected regions of the leaf as important features for some -- butnot all -- of the correctly classified images. For some images, the output ofthe interpretability methods indicated that spurious feature correlations mayhave been used to correctly classify them. Although the output explanation mapsof these interpretability methods may be different from each other for a givenimage, we advocate the use of these interpretability methods as `hypothesisgeneration' mechanisms that can drive scientific insight.

Usefulness of interpretability methods to explain deep learning based plant stress phenotyping