A Deep Learning-Based Approach for the Detection of Infested Soybean Leaves
Niklas Farah, Nicolas Drack, Hannah Dawel, Ricardo Buettner
Abstract
We address the soybean leaves infection problem by proposing a robust classification model that can reliably detect infests by Diabrotica speciosa and caterpillars. Our transfer-learning based model uses a VGG19 convolutional neural network to classify the soybean leaves and we achieve balanced accuracies between 93.71%and 94.16%on unseen testing data, what sets a new benchmark and outperform previous work using the same dataset. Our work has theoretical and practical implications. The soybean plays a crucial role in the agricultural industry. Infestation of soybeans leads to enormous economic and environmental losses. With our model presented here, an early and accurate detection to control the spread of plant pests is possible, what reduces economic and ecological damages.