Publication

Abstract : The greatest cause of death worldwide each year is mosquitoes. Identifying them is vital in order to take appropriate action to eradicate them in a particular location. Our aim is to create a state-of-the-art machine learning model to accurately identify and classify mosquitoes on human skin. This task is crucial in the field of public health as it can help to identify potential disease vectors and facilitate the implementation of prevention and control measures. We explored various pre-trained and deep convolutional neural network (DCNN) models for classification and evaluated the impact of hyperparameter tuning using the Hyperband optimization strategy. We also conducted preprocessing experiments and found data augmentation to be necessary. Our results demonstrated that both DCNNs and pre-trained models were effective in classifying mosquito species on human skin with high accuracy and F1 scores. We proposed an automated model update and build based on input images, which can be established in real-time environments by automating the hyperparameter selection on the best configuration. The use of the Hyperband optimization strategy was effective in improving model performance, with a significant increase in accuracy and F1 scores compared to models that were not tuned using Hyperband. By automating the process of selecting optimal hyperparameters for the pre-trained models using the Hyperband optimization strategy, we improved their performance on the classification of mosquito species on human skin, resulting in the best-performing result of this work with an accuracy of 91%. Our study provides valuable insights into the development of an artificial intelligence-based model for the termination of harmful mosquitoes through categorization. This work has important implications for the prevention and control of vector-borne diseases and the identification of potential vectors.