BioVision: AI Disease Segmentation focuses on leveraging deep learning models for semantic segmentation of diseases in chest X-rays. By utilizing advanced U-Net and U-Net++ architectures, this project aims to assist in early diagnosis and improved treatment planning.
Key Highlights
Dataset: The ChestX-Det dataset was converted to COCO format, enabling model training on true masks for accurate segmentation.
Models Used: U-Net and U-Net++.
Performance: Achieved mean IoUs of 0.88 (U-Net) and 0.86 (U-Net++) across all classes.
