Motion Retargeting with Single Frame

This paper aims to solve the video motion retargeting problem, where we are given a source video and target video, and we want to control the target person’s motion by the source person. While previous methods all require a minutes-long source video, we solve this problem with only one source frame, which greatly relaxes the data constraint of performing motion retargeting. To the best of our knowledge, our work is the first to approach motion retargeting with a single source frame. Experimental results demonstrate that our model is able to achieve the retargeting on the intermediate semantic maps, which is promising for future research on the realistic frames.