Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
This paper received good scores overall, with little disparity in the final scores. It extends HER to work with visual observations using a GAN to generate samples. The contribution relative to HER is clear. The main weaknesses of this work are lack of details about the generative model training, scalability concerns raised by reviewers and reliance on human annotated frames containing the goal.