Yunlong Liu

Efficiently fusing spatial and spectral information with CSI-Net leads to significantly improved change detection performance in remote sensing images.

MSI-Net not only outperforms existing change detection techniques but also addresses the unique challenges of post-earthquake imagery, paving the way for faster disaster response.

Robots can now learn fine-grained manipulation skills directly from human demonstrations, thanks to a new imitation learning framework that automatically figures out the right rewards and handles noisy real-world data.