Breast Cancer Detection with CNN

This was a focused medical-imaging classifier built around transfer learning rather than a from-scratch model. The goal was to compare how well established CNN architectures could adapt to the dataset under a constrained training setup.

What I built

• Trained a CNN model for breast cancer image analysis and reached 82% detection accuracy.
• Applied transfer learning with EfficientNet and Inception-ResNet-v2, improving performance by 20%.
• Used PyTorch for model work and Python data tooling for preprocessing, evaluation, and experiment tracking.