Detection of Mosquito Larvae Using Convolutional Neural Network

Abstract

Mosquito is responsible for the transmission of world's deadly diseases. Although a great deal of equipment are available to get rid of this species, most consist of creating barriers or killing adult mosquitoes causing very little effect on its population. This paper proposes a different approach of mosquito control using computer vision to detect the presence of early stage of mosquito's lifecycle. This paper introduces a Convolutional Neural Network based model that can detect the presence of mosquito larvae in water. The model can be used on both personal computers and small portable onboard computers like Raspberry Pi for portable or robotics application. Therefore, the model can be used on a system for the detection of mosquito larvae in order to eliminate its habitat. The model was developed from scratch and obtained a training accuracy of 93.95% and testing accuracy of 90.18%. Further testing with 100 images, the accuracy was found to be 86.0% and precision was 92.2%.