BasketballGAN:Generating Basketball Play Simulation Through Sketching
碩士 === 國立交通大學 === 多媒體工程研究所 === 107 === This thesis presents a data-driven basketball set play simulation, generated with a conditional adversarial network framework, that imitates different scenarios that may occur given an offensive set play sketch, leading to newly discovered insight by coaches an...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | en_US |
| Published: |
2019
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/x88e6u |
| Summary: | 碩士 === 國立交通大學 === 多媒體工程研究所 === 107 === This thesis presents a data-driven basketball set play simulation, generated with a conditional adversarial network framework, that imitates different scenarios that may occur given an offensive set play sketch, leading to newly discovered insight by coaches and players on how a given set play can be executed. By utilizing NBA movement data, we can train our generative model to imitate the behaviors of how players move around the court through two major components: a generator that learns to generate natural player movements given a latent noise and user sketched set play; and three discriminators that is used to evaluate the realism of two local variables (Offence and defence) and one global variable(Offence and defence combined). To establish the ball and player relationship, we introduce a dribblers penalty that integrates a ball and player relationship,prevents the ball from drifting away from the dribbler/ball handler or intended receiver from a pass.
|
|---|