A Solution for Counting Aedes aegypti and Aedes albopictus Eggs in Paddles from Ovitraps Using Deep Learning
Keywords:
Aedes aegypti, Aedes albopictus, ovitraps, egg counting, deep learningAbstract
The World Health Organization, in 2014, estimated that around one billion people get infected and one million people die from mosquito-borne diseases annually. The Aedes aegypti and Aedes albopictus are two of the species of mosquitoes which are transmission vector of diseases such as dengue, yellow fever, and chikungunya. Hence, monitor and control the population of these insects is a key factor to reduce the number of infections and deaths caused by them. One step towards the control of the population of mosquitoes is the usage of Ovitraps which is, basically, a dark container filled with water and with a porous wooden paddle where mosquitoes can lay their eggs. These devices are installed in a monitored area and periodically technicians collect them to manually count the number of eggs deposited in the paddles. Since the manual egg counting task can be time-consuming and human error can be introduced, in this work, we propose a complete solution (software and hardware) that scans the paddle loaded into it, detects and counts the number of eggs, and report it to the technician. The results achieved shows that the hardware was able to collect images with adequate quality and the selected model was able to detect the eggs with an average detection accuracy of 91%. Compared with two other tools for counting Aedes eggs in images, ICount and EggCounter, the proposed method achieved superior results. Furthermore, the Wilcoxon test indicates, with a confidence interval of 95%, that the presented solution can be as accurate as of the manual counting method which is currently adopted.