Characterisation of features in x-ray images of mice using deep learning

Class activation maps from classification convolutional neural networks are used to investigate features of the mouse skeletal system that can be used to determine the sex of mice. This work will help with understanding genotype-phenotype effects, in particular sexual dimorphism (the presentation of different non-sexual organ characteristics by two sexes of the same species). Preliminary results have been presented at the IEEE CyberNigeria 2021 conference.

Automatic classification of DSS ( dextran sulfate sodium ) histology images using deep learning

This project uses convolutional neural networks to predict the scores attributed to damage present in histology images of the mouse gut using caused by dextran sulfate sodium. Automating the scoring process for these images will help in drug and disease research.

Applying machine learning to paperless software pipelines

We are working with an industrial partner to incorporate machine learning into their paperless software pipelines. In particular we are recognising images and text in documents and automatically translating these to html forms to facillitate accurate and fast conversion of paper-based forms to digital equivalents.

Automatic construction of 3D avatars

This is a collaboration with a major Nigerian animation studio. We are applying state of the art techniques in 3-dimensional facial recognition and motion capture to construct 3D avatars that can be used for animation purposes. This exciting project will decrease the turnaround time for the studio to produce animations as well as reduce cost and free up valuable human resources that will otherwise be involved in this task.