Spectrally Distilled Representations Aligned with Instruction-Augmented LLMs for Satellite Imagery

Minh Kha Do, Wei Xiang, Kang Han +5

Vision-language foundation models (VLFMs) promise zero-shot and retrieval understanding for Earth observation. While operational satellite systems often lack full multi-spectral co…