首页|New Findings from University of Minnesota in the Area of Robotics Described (Kin et: Unsupervised Forward Models for Robotic Pushing Manipulation)
New Findings from University of Minnesota in the Area of Robotics Described (Kin et: Unsupervised Forward Models for Robotic Pushing Manipulation)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News-Current study results on Robotics have been published. According to news originating from Minneapolis, Minnesota, by N ewsRx correspondents, research stated, "Centric representation is an essential a bstraction for forward prediction. Most existing forward models learn this repre sentation through extensive supervision (e.g., object class and bounding box) al though such ground-truth information is not readily accessible in reality." Financial supporters for this research include Sony Research Award Program, Nati onal Science Foundation (NSF). Our news journalists obtained a quote from the research from the University of M innesota, "To address this, we introduce KINet (Keypoint Interaction Network)-an end-to-end unsupervised framework to reason about object interactions based on a keypoint representation. Using visual observations, our model learns to associ ate objects with keypoint coordinates and discovers a graph representation of th e system as a set of keypoint embed dings and their relations. It then learns an action-conditioned forward model using contrastive estimation to predict future keypoint states. By learning to perform physical reasoning in the keypoint spac e, our model automatically generalizes to scenarios with a different number of o bjects, novel backgrounds, and unseen object geometries."
MinneapolisMinnesotaUnited StatesN orth and Central AmericaEmerging TechnologiesMachine LearningRoboticsRob otsUniversity of Minnesota
NETL
NSTL
