Not All Patches are Equal: Sampling Matters for Visible-Infrared Pre-Training
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 be unreliable in VIS-IR data because imaging-physics differences make some spatially paired regions inherently less comparable, and aligning them with equal strength hinders representation learning and downstream transfer. In this paper, we revisit VIS-IR pre-training from a sampling perspective and propose Importance-Aware Sampling (IAS), which adjusts training emphasis based on patch reliability. Specifically, IAS (
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- LinkedLinked via arxiv author · 85%Qiwei Ma →
“Not All Patches are Equal: Sampling Matters for Visible-Infrared Pre-Training”
- LinkedLinked via arxiv author · 85%Bin Deng →
“Not All Patches are Equal: Sampling Matters for Visible-Infrared Pre-Training”
- LinkedLinked via arxiv author · 85%Junjie Zhu →
“Not All Patches are Equal: Sampling Matters for Visible-Infrared Pre-Training”
- LinkedLinked via arxiv author · 85%Qiangjuan Huang →
“Not All Patches are Equal: Sampling Matters for Visible-Infrared Pre-Training”
- LinkedLinked via arxiv author · 85%Puhong Duan →
“Not All Patches are Equal: Sampling Matters for Visible-Infrared Pre-Training”
- LinkedLinked via arxiv author · 85%Ke Yang →
“Not All Patches are Equal: Sampling Matters for Visible-Infrared Pre-Training”
- LinkedLinked via arxiv author · 85%Xudong Kang →
“Not All Patches are Equal: Sampling Matters for Visible-Infrared Pre-Training”
- LinkedLinked via arxiv author · 85%Shutao Li →
“Not All Patches are Equal: Sampling Matters for Visible-Infrared Pre-Training”