MotionRL: Align Text-to-Motion Generation to Human Preferences with Multi-Reward Reinforcement Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liu, Xiaoyang, Mao, Yunyao, Zhou, Wengang, Li, Houqiang
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910641725374464
author Liu, Xiaoyang
Mao, Yunyao
Zhou, Wengang
Li, Houqiang
author_facet Liu, Xiaoyang
Mao, Yunyao
Zhou, Wengang
Li, Houqiang
contents We introduce MotionRL, the first approach to utilize Multi-Reward Reinforcement Learning (RL) for optimizing text-to-motion generation tasks and aligning them with human preferences. Previous works focused on improving numerical performance metrics on the given datasets, often neglecting the variability and subjectivity of human feedback. In contrast, our novel approach uses reinforcement learning to fine-tune the motion generator based on human preferences prior knowledge of the human perception model, allowing it to generate motions that better align human preferences. In addition, MotionRL introduces a novel multi-objective optimization strategy to approximate Pareto optimality between text adherence, motion quality, and human preferences. Extensive experiments and user studies demonstrate that MotionRL not only allows control over the generated results across different objectives but also significantly enhances performance across these metrics compared to other algorithms.
format Preprint
id arxiv_https___arxiv_org_abs_2410_06513
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle MotionRL: Align Text-to-Motion Generation to Human Preferences with Multi-Reward Reinforcement Learning
Liu, Xiaoyang
Mao, Yunyao
Zhou, Wengang
Li, Houqiang
Computer Vision and Pattern Recognition
We introduce MotionRL, the first approach to utilize Multi-Reward Reinforcement Learning (RL) for optimizing text-to-motion generation tasks and aligning them with human preferences. Previous works focused on improving numerical performance metrics on the given datasets, often neglecting the variability and subjectivity of human feedback. In contrast, our novel approach uses reinforcement learning to fine-tune the motion generator based on human preferences prior knowledge of the human perception model, allowing it to generate motions that better align human preferences. In addition, MotionRL introduces a novel multi-objective optimization strategy to approximate Pareto optimality between text adherence, motion quality, and human preferences. Extensive experiments and user studies demonstrate that MotionRL not only allows control over the generated results across different objectives but also significantly enhances performance across these metrics compared to other algorithms.
title MotionRL: Align Text-to-Motion Generation to Human Preferences with Multi-Reward Reinforcement Learning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2410.06513