Self-supervision assisted multimodal remote sensing image classification with coupled self-looping convolution networks

Shivam Pande, Biplab Banerjee

Abstract: Highlights•Self-looping blocks with shared weights across time for feature refinement.•Multiscale kernels are used for self-looping blocks for spatial diversity.•Cross-modal communication with shared weights across modalities.•Weight sharing across modalities and time enable efficient parameter controlling.•Robust cross-modal learning with self-supervised cross-modal reconstruction.