多智能体协同感知

基于协同视角的无监督多智能体与单智能体感知

Unsupervised Multi-agent and Single-agent Perception from Cooperative Views

利用多智能体协同视角的点云密度增益与跨视角一致性作为无监督信号，为单车单视角 3D 检测生成伪标签，可直接迁移到车路协同、V2X 等无标注部署场景。