This markdown will provide a step-by-step guide on how to replicate the result of my COMP4093 Thesis project.

Make sure you have python and anaconda set up to be able to run the notebooks. You can download the Anaconda distribution from here or use whatever package distribution you have (winget, brew, apt, etc.).

Dataset prep:

You can either skip this and use the finished dataset in splited-stratified or follow the steps below to replicate the creation of the dataset.

Labeling:

1. Set up Label Studio then import the dataset from 'raw' folder.

2. Label the dataset based on the framework stated in the methodology section of the thesis.

3. Extract it so we can use it in the next step.

Stratified and Splitting:

1. Run the notebook aug.ipynb in data-augmentation folder to augment the dataset. Output should be saved in /dataset/augmented.

2. Run the notebook stratified-split.ipynb in /dataset/splited-stratified to split the dataset into training, validation, and test sets. The output will be saved in /dataset/splited-stratified.

Fine-tuning and result collection:

1. Run each of the notebook from their respective model folder. Individual results will be saved in the /results folder.

2. Open the results folder then run shared-diagram.ipynb to generate the model comparison diagram.