This library contains various scripts useful for running machine learning projects, including performing hyperparameter optimization and k-fold cross validation while storing results with WandB. Some work is currently under development.
Author's Note: Some projects associated with the Master's Thesis are published on Google Scholar. https://scholar.google.com/citations?user=SJCymuoAAAAJ&hl=en
Author Profile: https://www.linkedin.com/in/anish-s-36179a97/
KCV_HP_Template: A template script which can perform k-fold cross validation & hyperparameter optimization
MS Thesis: Machine Learning for Abdominal Aortic Aneurysm Characterization from Standard-Of-Care Computed Tomography Angiography Images
Masters_Thesis: Notebooks associated with the completion of the Master's Thesis
1. AAA-UNet: Baseline Aneurysm Segmentation
2. BB-AAA-UNet: Memory Efficient High-Resolution Segmentation with Prior Aneurysm Localization
3. BB-AAA-UNet: As Applied to Aneurysm Wall Segmentation
4. Patch Segmentation UNet: Prediction of Aneurysm Wall by Medical Image Sub-volumes
5. AAA Image Transformers: Classifying Medical Images by Aneurysm Severity with Latent Representations
6. AAA-ViT: Moving Towards Detection with Classification of Aneurysm Severity with Anatomical Explanation
Peripheral Artery Disease Classification from Computed Tomography Angiography Images via 3D Medical Image Vision Transformers with Explainability
PAD_ViT: Repository of a medical image classification project for ImageRx.
Any Segmentation Model: The 3D Foundational Segmentation Model to Revolutionize Medical Image Annotation
ASM: An example use case of applying the 3D Foundational Model to segment volumetric data. This model was outfitted with a text encoder. Based on: https://github.com/facebookresearch/segment-anything
Self-Supervised Medical Image Classification of Radiographs via Convolutional Neural Network Inpainting and Class Balanced Loss Functions
Self_Supervised_Learning: An example of self-supervised learning in action.
MEDVQA: A vision language model.
RetinaNet: An automatic face mask detector which uses bounding boxes. The script is the one stop shop for data curation, model development, statistical benchmarking, and deployment on a local computer.
Coding_Questions: A folder of various quick and simple machine learning scripts for practice.