Parameter-efficient fine-tuning for single image snow removal

Xinwei Dai, Yuanbo Zhou, Xintao Qiu, Hui Tang, Tong Tong

Abstract: Highlights•First to bring efficient parameter fine-tuning to low level vision tasks.•Proposing a method to explore features from large models for small models.•Surpassing previous methods with just 15% of parameters trained.•A plug-and-play framework to enhance the performance of UNet in restoration tasks.