ameliacode/character-boxing

Character Boxing in GAN-based Reinforcement Learning

Character Boxing in GAN-based methods

Code from "Exploring the Effectiveness of GAN-based approach and Reinforcement Learning in Character Boxing Task"

Look for papers here papers will be out soon

---

Installation

out downstream tasks are based on ASE and AMP.
Clone ASE repository first, then follow the instructions from its repo. To execute Multi Agent RL environment, clone TimeChamber.

Table down below is required setting for each project. Using virtual env is recommended

Project	Python	RL-games
AMP/ASE(punching, joystick)	3.7	1.1.4
TimeChamber(boxing)	3.7/3.8	1.5.2

Preparation

Our repository only consists of task files.
To train from scratch, see Preprocessing section.

Preprocessing

All motion data used were from CMU We used data that were labeled as boxing and locomotion.