PeftCD: Leveraging Vision Foundation Models With Parameter-Efficient Fine-Tuning for Remote Sensing Change Detection

To tackle the prevalence of spurious changes, the scarcity of annotations, and the difficulty of cross-domain transfer in multi-temporal and multi-source remote sensing imagery, we proposePeftCD, a change detection framework built uponVision Foundation Models (VFMs)withParameter-Efficient Fine-Tuning (PEFT). Specifically, PeftCD adopts a shared-weights Siamese encoder instantiated from a VFM, into whichLoRAandAdaptermodules are injected as fine-tuning strategies, so that only a small number of additional parameters need to be trained for task adaptation. To better explore the potential of VFMs in change detection, we investigate two representative backbones: theSegment Anything Model v2 (SAM2), which provides strong segmentation priors, andDINOv3, a state-of-the-art self-supervised representation learner. Meanwhile, PeftCD employs a deliberatelyminimal and efficient decoderto highlight the representational capacity of the backbone models. Extensive experiments demonstrate that PeftCD achievesstate-of-the-art performanceacross multiple public datasets, including SYSU-CD (IoU 73.81%), WHUCD (92.05%), MSRSCD (64.07%), MLCD (76.89%), CDD (97.01%), S2Looking (52.25%) and LEVIR-CD (85.62%), with notably precise boundary delineation and strong suppression of pseudo-changes. Overall, PeftCD achieves a favorable balance among accuracy, efficiency, and generalization, offering an efficient paradigm for adapting VFMs to practical remote sensing change detection. The code and pretrained weights are available athttps://github.com/dyzy41/PeftCD.