metadata
task_categories:
- image-segmentation
license: mit
This repository contains the medical imaging datasets used in SALT: Parameter-Efficient Fine-Tuning via Singular Value Adaptation with Low-Rank Transformation. SALT is a novel Parameter-Efficient Fine-Tuning (PEFT) method designed to adapt large-scale foundation models—especially **Segment Anything Model (SAM)**—to domain-specific tasks such as medical image segmentation.
The datasets are:
- ROSE (Retinal OCT Angiography)
- ARCADE (Coronary Artery Segmentation)
- DRIVE (Retinal Vessel Segmentation)
- DIAS (Dynamic Digital Subtraction Angiography)
- Xray-Angio (Occluded Vessel Segmentation)