Emmanuel Denu
PIPS Host Organisation: UK Health Security Agency (UKHSA)

Emmanuel Denu undertook a project at the UK Health Security Agency (UKHSA) to develop a deep learning mosquito image recognition model aimed at detecting species invasive to the United Kingdom. The project involved three stages: an insect detector model, a binary classifier, and a species classifier.

The insect detector model was designed to crop out insects within photos, which were then analyzed by the binary classifier to determine if they were mosquitoes. If identified as mosquitoes, the images were further processed by the species classifier to identify the specific species. Emmanuel contributed by creating synthetic mosquito images for the binary classification and insect detector models, utilizing existing images from UKHSA databases. He successfully trained a binary classifier using mosquito and non-mosquito images and developed a species classifier for 11 mosquito species. Additionally, he created a program to access these models and extract misclassified images for further assessment by the medical entomology team.

Throughout the placement, Emmanuel mastered project collaboration and version control using Git and GitHub, as well as project management and tracking with GitHub tickets. He gained experience working in a team, sharing responsibilities, and achieving set goals. The project provided a platform for him to apply his skills in building and training deep learning models in an industry setting, and he also honed his ability to write and organize code for reusability.

Emmanuel’s proudest moment was presenting the project outcome to the Medical Entomology and Zoonosis Ecology team, receiving positive feedback and impressing his team with the progress made since joining them.