19 / 2023-08-23 12:51:33
Human Motion Prediction Based on Graph Convolutional Networks and Multilayer Perceptron
Human motion prediction, graph convolutional networks, Time information, Spatial dependence, discrete cosine transform
终稿
Ziliang Ren / Dongguan University of Technology
Jin miaomiao / Dongguan University of technology
In the field of computer vision, human motion prediction is a classic problem with many great uses. Methods for predicting human movement aim to process and analyze human movement data for predicting future human movement. There are complex problems of loss constraint and training process in this method. For solving the above problems, reduce the training and prediction time and provide a new perspective, this paper uses a lightweight and efficient graph convolutional network combined with multi-layer perceptron model, including a fully connected layer graph convolutional network. What’s more, for further improving the prediction accuracy, we also compare some paper models with this model and find that this model has better accuracy performance. Finally, experiments on Human3.6M dataset verify the effectiveness of the proposed method, the proposed method can accurately predict the future human behavior, the accuracy can reach 90.2%.In the field of computer vision, human motion prediction is a classic problem with many great uses. Methods for predicting human movement aim to process and analyze human movement data for predicting future human movement. There are complex problems of loss constraint and training process in this method. For solving the above problems, reduce the training and prediction time and provide a new perspective, this paper uses a lightweight and efficient graph convolutional network combined with multi-layer perceptron model, including a fully connected layer graph convolutional network. What’s more, for further improving the prediction accuracy, we also compare some paper models with this model and find that this model has better accuracy performance. Finally, experiments on Human3.6M dataset verify the effectiveness of the proposed method, the proposed method can accurately predict the future human behavior, the accuracy can reach 90.2%.
重要日期
  • 会议日期

    11月02日

    2023

    11月04日

    2023

  • 12月15日 2023

    初稿截稿日期

  • 12月20日 2023

    注册截止日期

主办单位
IEEE Instrumentation and Measurement Society
Xidian University
移动端
在手机上打开
小程序
打开微信小程序
客服
扫码或点此咨询