T-SAR-JEPA: Self-supervised anomaly detection in SAR images

Arxiv CS.CV (computer vision)

T-SAR-JEPA adapts a ViT-Base/16 encoder on 39,300 Capella patches. It performs temporal anomaly detection in SAR stacks via latent prediction. The self-supervised framework uses local masked reconstruction.

self-supervised-learningvision-transformer

Arxiv CS.CV (computer vision)·Kerod WoldesenbetAbem Woldesenbet

6 days ago