metadata
task_categories:
- image-segmentation
license: mit
tags:
- parameter-efficient-fine-tuning
- peft
- segment-anything
- sam
- medical-image-segmentation
This repository contains the code for SALT: Parameter-Efficient Fine-Tuning via Singular Value Adaptation with Low-Rank Transformation. SALT is a method for adapting large-scale foundation models, particularly the Segment Anything Model (SAM), to domain-specific tasks, such as medical image segmentation, with high parameter efficiency.
Paper: SALT: Parameter-Efficient Fine-Tuning via Singular Value Adaptation with Low-Rank Transformation
Code: https://github.com/YourUsername/SALT.git
The following datasets, used in the experiments, are available on Hugging Face:
- ROSE: (https://huggingface.co/datasets/pythn/ROSE)
- ARCADE: (https://huggingface.co/datasets/pythn/ARCADE)
- DRIVE: (https://huggingface.co/datasets/pythn/drive)
- DIAS: (https://huggingface.co/datasets/pythn/DIAS)
- Xray-Angio: (https://huggingface.co/datasets/pythn/DB)