A novel approach to identifying Mosquito Breeding Grounds with Convolutional Neural Networks, to stop the spread of Vector-Borne Diseases.

Mosquitoes are the deadliest animal on Earth, responsible for over 725,000 deaths annually due to the diseases they spread. Malaria deaths alone increased by 82% from 1980 to 2004, highlighting a growing global health crisis. To combat this problem, I developed two custom Convolutional Neural Networks (CNNs) based on the Ultralytics You Only Look Once (YOLO) v8 standard to detect Mosquito Breeding Grounds (MBGs) in both aerial and ground-based imagery. Proactively identifying and removing the MBGs promptly can save lives. I trained my aerial model, which has three million parameters, 225 layers, and a mean Average Precision at 50 Percent Intersection over Union score (mAP@50IoU) of over 90%, on a dataset comprising over 5,100 aerial images of MBGs. My UpClose model has the same structure, a dataset of over 1,200 photographs, and took 225 Epochs to train. My Aerial model can enhance the precision and efficiency of public health efforts by enabling administrative bodies worldwide to detect MBGs via drones and other aerial methods, allowing for their prompt removal. My UpClose model empowers everyday citizens to make a positive impact in their communities by detecting MBGs using a smartphone. This project demonstrates the transformative potential of Machine Learning in identifying MBGs with high accuracy, preventing outbreaks, and saving thousands of lives, making a global health impact.