SAFERNet: Channel, Positional, and Global Attention Fusion for Efficient RGB-T Segmentation in Disaster Robotics | Publicación