WingBeats - A Deep Learning-based Mosquito Species and Gender Classification System
Project Members
- Prof. Dr. Peter Haddawy, Faculty of ICT, Mahidol University
- Asst. Prof. Dr. Patchara Siriwichai, Faculty of Tropical Medicine, Mahidol University
- Dr. Myat Su Yin, Faculty of ICT, Mahidol University
- Chaitawat Sa-ngamuang, Faculty of ICT, Mahidol University
- Chanaporn Chaisumritchoke, Faculty of ICT, Mahidol University
- Tup kongthaworn, Faculty of ICT, Mahidol University
- Borvorntat Nirandmongkol, Faculty of ICT, Mahidol University
Diseases transmitted by mosquitoes such as malaria, dengue, West Nile fever, and most recently Zika Fever are amongst the biggest healthcare concerns across the globe today. In Thailand, Aedes. aegypti, Ae. albopictus are the primary vectors of dengue and Annopheles. dirus, An. minimus, the primary vectors for malaria. To tackle such life-threatening diseases, it is vital to evaluate the risk of transmission. Of critical importance in this task is the identification of vector species prevalent in an area of interest. In addition, identifying gender is important since only females of the vector species transmit disease. Traditional approaches to identify the species and gender are both time and labor-intensive to collect specimens and identify the data by using trained personnel. We will implement a web-based system for automatic classification mosquito species to identify the species and gender of mosquito from audio recordings wingbeat sounds.
In our data collection, we will consider various conditions that could affect the wingbeat frequency for example age, quality of the microphone, group size, and species mix. Techniques from signal processing will be used to characterize the wingbeat patterns before building state-of-the-art deep learning-based classification models. The models will be evaluated using the pre-recorded mosquitoes as well as the recording from mosquitoes in a live session in a laboratory setting.