SegDAC: Improving Visual Reinforcement Learning by Extracting Dynamic Object-Centric Representations from Pretrained Vision Models

VLMRe-assign community