Sheet 04 / Research

Research

Machine learning and computer vision, mostly from university. I care a lot about how results are measured: a model that only looks good on a lucky test split isn't much use.

04.1

MSci dissertation

Human Activity Recognition on WISDM 2.0

A comparison of classical machine-learning and deep-learning approaches to smartphone-based Human Activity Recognition, using user-independent evaluation.

Gradient Boosting

0.981

LSTM

0.980

LSTM-CNN

0.958

0.950.960.970.98

Fig. 1 — Macro F1, user-independent split. Axis starts at 0.95.

Read the case study →

04.2

Education

University of Exeter

MSci Computer Science (Integrated Master's)

First Class (Hons) · September 2021 – June 2025

Machine Learning91%
Enterprise Computing83%
Computer Vision81%
Group Development Project80%
Artificial Intelligence77%

First-class threshold: 70%

04.3

Applied work

01Automated Race Timing SystemComputer vision that spots runners, reads their race bibs and times them automatically.ResearchCompleted
02Human Activity Recognition (WISDM 2.0)Classical machine learning versus deep learning for recognising activity from smartphone sensors.Research · 2025Completed research
03Cycling FTP PredictionPredicting a cyclist's Functional Threshold Power from their training data.ResearchPrevious project