Research to synthesize tennis matches in a realistic way. Based on the domain knowledge of tennis, they divided the player's movement into two parts, “catching up and hitting the ball" and "taking the next position”. They select the nearest frames from them to synthesize. There are no shadows, so it can be identified as a synthesized, but the movements themselves are quite natural.
TL;DR
Research to synthesize tennis matches in a realistic way. Based on the domain knowledge of tennis, they divided the player's movement into two parts, “catching up and hitting the ball" and "taking the next position”. They select the nearest frames from them to synthesize. There are no shadows, so it can be identified as a synthesized, but the movements themselves are quite natural.
Why it matters:
Paper URL
https://arxiv.org/abs/2008.04524
Submission Dates(yyyy/mm/dd)
2020/08/11
Authors and institutions
Haotian Zhang, Cristobal Sciutto, Maneesh Agrawala, Kayvon Fatahalian
Methods
Results
Comments
project page : https://cs.stanford.edu/~haotianz/research/vid2player/