Age and species prediction of natural malaria mosquitoes through deep learning of mid-infrared spectra

Josh Mitton, 2 Feb 2020

This is a summary of work completed so far

Here we use a 1-D deep convolutional neural network architecture to predict both age and species of mosquitoes given their mid-infrared spectra. In the case of lab reared mosquitoes, where a large number of samples exist, this modelling technique perfroms strongly.
Age Prediction of Lab Reared Lab Mosquitoes
Species Prediction of Lab Reared Lab Mosquitoes
Age Prediction of Lab Reared Field Mosquitoes
Species Prediction of Lab Reared Field Mosquitoes
When attempting to learn on and make predictions of mosquitoes reared in a semi-field the model does not perform as strongly, with one possible cause being the lack of semi-field reared mosquito data. In the paper we develop methods to overcome this, yielding a high classification accuracy with a relatively low number of semi-field reared mosquito samples.
Predicting Age and Species of Semi-Field Reared Mosquitoes