Aspiring Developer (AI/ML or Web)

Artefact Classification

This project was created and put together for my dissertation as part of my MSc in Computer Science. It involved the creation of a novel cultural-heritage specific dataset which covers 6 classes with a combined total of 223,486 images. The quality of this dataset was demonstrated using various CNNs pre-trained on ImageNet. ResNet50 and EfficientNetB0 to B2 were the models chosen for this task and all achieved between 84% and 86% global accuracy. Additionally, we investigated how much class imbalance can impact the accuracy of a CNN using class weights, undersampling, and oversampling techniques. Our experiments showed that undersampling and class weights made our models worse. In contrast, oversampling using data augmentation helped our model be more balanced by bringing its precision and recall values closer together; they attained 68.9333% and 63.5916% respectively. As there is very little research in art classification and no cultural-heritage specific dataset readily available for visual classification tasks, the findings and output of this project opened up new avenues of research for cultural-heritage tasks.

The Traveling Salesman Dilemna

The Traveling Salesman problem is one of the most famous example of NP problems : "Given a list of cities and the distances between each pair of cities, what is the shortest possible route that visits each city exactly once and returns to the origin city?" In this project, we created a genetic algorithm in order to find a solution deemed 'good enough' to this problem. There are five stages to this algorithm : (1) Initialisation of the population, (2) Fitness function, (3) Selection of the fittest individuals (the parents), (4) Reproduction, (5) Mutation. In this project, the Tournament Selection was chosen as the selection method. For reproduction, due to the nature of this NP-problem, Ordered Crossover was selected. Finally, for the mutation step, Swap Mutation was chosen.

Snake

Having grown up playing snake on a Nokia 3310, I wanted to pay homage to this classic game by implementing my own version in Python, using the Pygame library.

The above image for this project was found on hmd.com. All credits to them. Their article with said picture can be found : here

Online Portfolio

This online portfolio which you are currently browsing was created to act as an online resume showcasing all my acquired skills. This was also an opportunity to learn to work with Sass and reinforce my grip over HTML, CSS, and Boostrap.

Code First Girls - Course Project

Following a short 8-weeks course to learn the fundamentals of web development, our last task was to create a website (1-2 pages) to showcase all we had learnt. For this end of course project, I decided to put together a small blog (main page and contact page). The website is reponsive and behaves well on any device and screen size.