Onboard vision-language models could enable satellites to answer queries directly, but exhaustive tiled inference over high-resolution imagery is slow and en…
机构:UIUC
来源:arXiv 2609.29029 | AI4Papers 论文推荐平台