Small and Clustered Objects Counting — November 2021 – February 2024

• Collected, cleaned, and annotated a bespoke dataset to support model training.
• Developed two deep learning algorithms to count small and clustered objects with limited training data.
• Integrated a Transformer-based architecture to improve model generalisation.
• Achieved 9.62% MAPE on colony counting while eliminating manual counting effort.
• First-authored a paper (under review) and presented findings at journal clubs and local conferences.

Investigation of Colony Counting with XAI — March 2020 – October 2021

• Implemented the best-performing colony-cardinality classification model using PyTorch/TensorFlow.
• Applied Explainable AI (XAI) techniques to interpret and validate the classifier’s predictions.
• Generated synthetic colony images using Generative Adversarial Networks (GANs).
• Explored alternative loss functions to mitigate class imbalance.
• Submitted the findings as a journal paper, currently under review.

Text Recognition on Antibiotic Discs — March 2019 – September 2019

• Developed a computer vision algorithm to recognise text printed on antibiotic discs from petri dish images.
• Improved recognition accuracy to 95.2% while optimising processing speed.

Load Balancing Algorithms — May 2017 – September 2017

• Implemented Simulated Annealing and Genetic Algorithm approaches to analyse cloud computing load balancing.
• Benchmarked both algorithms against a brute-force baseline.
• Co-authored a paper presenting the findings.

See the Publications page for the papers arising from this work.