Back to Projects
LivePublic Projectopen-source
Cataract AI
Deep learning system for automated cataract detection
AI Research EngineerDec 2025
About This Project
A medical AI research project focused on detecting cataracts from retinal fundus images. Using a ResNet50 architecture and transfer learning, the system addresses extreme pathological class imbalances to provide a high-sensitivity triage tool for healthcare providers.
Key Highlights
1
ResNet50-based transfer learning model for medical diagnosis
2
Implemented mathematically weighted loss functions to handle 93% class imbalance
3
Achieved high recall for critical medical triage accuracy
4
Automated image preprocessing pipeline for fundus photography
Technology Stack
PythonPyTorchOpenCVScikit-learn
