An Improved Transfer Learning Based Larvae Detection and Classification Using Densenet121

Bangladesh is currently experiencing a surge in dengue fatalities due to increased Aedes mosquito activity during rainy seasons. To address this issue, this paper proposes a solution for detecting Aedes larvae in water bodies near and inside homes using deep learning models. A pre-processed image dataset is utilized to train transfer learning models, including DenseNet-121, Inceptionv3, Xception, ResNet50, VGG16, VGG19, and Ensemble boosting. The results demonstrate the exceptional performance of the DenseNet121 model, achieving 100% accuracy, precision, and recall scores. Therefore, a system has been developed to assist social workers in classifying collected images and detecting Aedes and non-Aedes larvae. This study highlights the potential of using deep learning models to classify mosquito larvae, offering significant implications for controlling diseases transmitted by mosquitoes such as dengue and Zika. By leveraging technological advancements and community engagement, the prevention and management of mosquito-borne illnesses can be greatly improved.