Screen, Cache, and Match: A Training-Free Causality-Consistent Reference Frame Framework for Human Animation
屏幕、缓存与匹配:一种无需训练的因果一致性参考框架用于人类动画
机构 * College of Intelligent Robotics and Advanced Manufacturing, Fudan University(复旦大学智能机器人与先进制造学院) ; Shanghai Key Lab of Intelligent Information Processing, College of Computer Science and Artificial Intelligence, Fudan University(复旦大学计算机科学与人工智能学院上海市智能信息处理重点实验室) ; College of Design and Innovation, Tongji University(同济大学设计创意学院) ; School of Data Science and Engineering, East China Normal University(华东师范大学数据科学与工程学院)
AI总结 本文提出FrameCache框架,通过Screen-Cache-Match策略和Trajectory-Aware Autoregressive Generation机制,提升人类动画的时空一致性与视觉稳定性。