More than the Sum: Panorama-Language Models for Adverse Omni-Scenes

Weijia Fan, Ruiping Liu, Jiale Wei +7

Existing vision-language models (VLMs) are tailored for pinhole imagery, stitching multiple narrow field-of-view inputs to piece together a complete omni-scene understanding. Yet,…