1 paper
Qiwei Ma, Bin Deng, Junjie Zhu +5
Visible-infrared (VIS-IR) alignment is a key pre-training task for robust multi-sensor perception. Most existing methods use uniform patch-wise contrastive learning, but this can b…