Code for "Transfer Learning without Knowing: Reprogramming Black-box Machine Learning Models with Scarce Data and Limited Resources". (ICML 2020)

"Transformer-based end-to-end classification of variable-length volumetric data" that will appear in MICCAI 2023.

SCC-ALAR: An experimental study on skin lesion classification using hybrid CNN-Transformer architectures. This project explores the impact of architectural complexity versus inductive bias on the imbalanced HAM10000 dataset, featuring custom weighting schemes and pre-extracted feature embeddings.

ImFlow: A better image dataset loader for TensorFlow

Open-source glaucoma detection AI for mobile/low-resource clinics using synthetic training data

A modified Half-UNet model for brain tumor classification using MRI scans. This project repurposes the UNet architecture for classification by integrating dense layers at the bottleneck. Built with PyTorch and Albumentations, it demonstrates high accuracy on the Kaggle Brain Tumor Detection dataset.

An elegant macOS demo app built with SwiftUI that leverages CoreML for real-time bone fracture detection from X-ray images. This project demonstrates how deep learning models can be seamlessly integrated into Swift-based apps for medical image analysis and AI-powered diagnostics.

InceptionV3 skin disease classification (PyTorch) + Streamlit demo

All code created during master's thesis project at CFIN

Custom CNN built from scratch with TensorFlow for breast cancer classification. Optimized for single-GPU training using grayscale inputs, class balancing, and data augmentation. Includes hyperparameter tuning, metric validation, and Grad-CAM for explainability.

Hybrid ML pipeline using FCM segmentation, EfficientNetB4 feature extraction, and XGBoost for leukemia detection.

PyTorch implementation of ConvNeXt for Alzheimer's MRI classification (80.9% Accuracy).

Deep learning tool for Alzheimer’s stage classification from MRI scans using custom CNN (98% accuracy), saliency maps, confidence visualization, and Streamlit interface with PDF reports.

AI powered Image-Based application for detecting and predicting disease and abnormalities.

Deep Learning for Alzheimer’s Detection: Classifying anatomical MRI scans using advanced neural networks.

This project culminates in a model that can classify a given chest x-ray for the presence or absence of pneumonia.

Exploratory analysis and visualization of a breast cancer histopathology image dataset, including binary and multiclass classification structures across multiple magnification levels.

Classifies MRI scans into Non-Demented, Very Mild, Mild, and Moderate Dementia using a fine-tuned VGG16 model. Implements data augmentation, class weighting, and Grad-CAM for interpretability.