1 paper
Jialin Tian, Zhuofan Li, Xinkai Wang +5
Robot demonstrations record what the robot saw and what it did. They do not specify how the 3D world changed in between. Scene-level 3D flow describes it by tracking 3D scene point…