To make current artificial intelligence systems flexibly explore causal information between behaviors and understand events based on human language habits, we propose a hybrid 2D/3D convolutional network guided by human behavior, which aims to explore a new machine understanding video model by focusing on human activities.

中文翻译：

---

探索由行为引导的类人视频理解算法

为了使当前的人工智能系统能够灵活地探索行为之间的因果信息并基于人类语言习惯理解事件，我们提出了一种以人类行为为导向的混合 2D/3D 卷积网络，旨在通过关注人类活动探索一种新的机器理解视频模型.