This repository contains the work executed by me as an research intern at VIGIL labs of IIT Hyderabad.
The datasets used for the investigation were related to Brain Tumour Detection and can be downloaded from Kaggle using the link given below:
- source dataset : https://www.kaggle.com/datasets/ahmedhamada0/brain-tumor-detection
- target dataset : https://www.kaggle.com/datasets/navoneel/brain-mri-images-for-brain-tumor-detection
In particular, the following topics were studied and implemented for the above two mentioned datasets.
- Summary and Visualisation of the two datasets
- Implementing Grad-CAMs on the two datasets to extract relevant information from individual images.
- Using Autoencoder Average Distance (AAD) to find out similarities between the two datasets.
The notebook titled main.ipynb contains the above mentioned code and can be executed in Jupyter notebook or Google Colab by introducing minor changes. Interestingly, code given here is flexible and can be used for any two datasets with little changes.
In addition to these, following toics were also studied and implemented during course of internship ( their code is note given in this repository):
- Grad-CAMs on Office31 Dataset.
- Optimal Transport Theory to find out the distance between the two dataset.
- Direct Domain Adaptation to find out similarities between the two dataset.