As a wet-lab biologist, I always wondered — can a computer learn to see what we see under the microscope (the way a microscopist does)?
This repository is my exploration of that question — applying machine learning and CNN-based approaches to bioimage classification, starting from malaria cell detection and expanding to other biological phenotypes.
The goal is not just accuracy — but understanding HOW and WHY AI makes biological decisions, hence the focus on Explainable AI.
- Infected vs uninfected cell detection
- CNN architecture from scratch
- Evaluation: ROC curve, AUC, Confusion Matrix
- Explainable AI: GradCAM (under development)
- Dataset: Kaggle malaria cell images
- Normal vs apoptotic cell detection
- CNN + Explainable AI
- U-Net based segmentation
- Dataset: BBBC038
- Python, PyTorch
- Google Colab (Tesla GPU)
- scikit-learn
Nikhil Kirtipal
🔬 Always curious — from microscope to machine learning!