AI / Machine Learning
BUSI Breast Ultrasound Segmentation
A medical imaging project using U-Net and Attention U-Net architectures to segment breast ultrasound images from the BUSI dataset. The work spans model development and an interactive Streamlit deployment.
Problem solvedManual segmentation of medical scans is slow and inconsistent. This automates the first pass.
FeaturesTrain the models, run inference on new scans, and view the segmented mask output.
ChallengesGetting segmentation models to generalize well on a limited medical imaging dataset.
What I learnedAn end-to-end deep learning workflow, from data preparation and model training to an inference app.