Khoa Phan

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

Vision-language foundation models (VLFMs) promise zero-shot and retrieval understanding for Earth observation. While operational satellite systems often lack full multi-spectral coverage, making RGB-only inference highly desirable for scalable deployment, the adoption of VLFMs for satellite imagery