Mitigating the Human-Robot Domain Discrepancy in Visual Pre-training for Robotic Manipulation | Read Paper on Bytez