Back

Detection of insect health with deep learning on near-infrared sensor data

Bick, E. N.; Edwards, S.; De Fine Licht, H. H.

2021-11-16 ecology
10.1101/2021.11.15.468635 bioRxiv
Show abstract

Conventional monitoring methods for disease vectors, pollinators or agricultural pests require time-consuming trapping and identification of individual insects. Automated optical sensors that detect backscattered near-infrared modulations created by flying insects are increasingly used to identify and count live insects, but do not inform about the health status of individual insects. Here we show that deep learning in trained convolutional neural networks in conjunction with sensors is a promising emerging method to detect infected insects. Health status was correctly determined in 85.6% of cases as early as two days post infection with a fungal pathogen. The ability to monitor insect health in real-time potentially has wide-reaching implications for preserving pollinator biodiversity and the rapid assessment of disease carrying individuals in vector populations. One sentence summaryAutomated optical sensors distinguish between fungus-infected and healthy insects.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.