Learning Nonlinear Flow Matching for Multimodal Remote Sensing Image Clustering (opens in new tab)

Limited by different imaging mechanisms of sensors, multimodal remote sensing images hence form various data manifolds in observation spaces. Leveraging the multimodal complementarity is critical to improving clustering. In this article, we propose learning nonlinear flow matching (LNFM) that converts distributions of multimodal images to fuse multimodal features for clustering. Specifically, the nonlinear flow matching adaptively learns the nonlinear interpolation paths of the multimodal ima...