Speaker
Description
Mosquito borne diseases such as malaria, dengue, Zika virus and chikungunya impose a disproportionate heavy public health burden in Africa. While adult mosquitoes which are vectors of these diseases are widely studied, there is a persistent neglect in the developmental stages of the mosquitoes. Yet, these early stages are critical since vector control strategies often target larvae and pupae to break the transmission cycle before mosquitoes mature into biting adults. Moreover, the identification of these specimens is prone to human error by experts due to the striking similarity in morphological features between larval and pupal stages across species. This limitation not only slows down surveillance efforts but also increases the risk of misclassification, which can compromise vector control interventions. In response, we present an Africa-based curated dataset of the larval and pupal stages of Aedes, Anopheles, and Culex species, alongside non-mosquito specimens with similar morphology. This directly addresses the African dataset representation gap in global repositories and uniquely contributes to pupal-stage classification, which remains highly underrepresented in most classification tasks. We present MosqMixerNet, a lightweight deep learning model tailored for image-based classification of mosquito early-stages. The proposed MosqMixerNet balances compactness and accuracy with only 169,666 trainable parameters and was benchmarked against MobileNetV2, DenseNet121, and NASNetMobile. On genus level identification with a non-mosquito class, it reached 99.37% accuracy, outperforming NASNetMobile at 97.30%, MobileNetV2 at 97.26%, and DenseNet121 at 80.36%. The work advances computer vision for public health and tropical disease control, and provides a foundation for integration into digital vector monitoring platforms in endemic settings.