Title:
３Ｄ畳み込みニューラルネットワークに基づく動作識別方法及び装置
Document Type and Number:
Japanese Patent JP6920771
Kind Code:
B2
Abstract:
The present disclosure relates to a neural network-based identification method and device, the method comprising: inputting a video to be identified into a trained first three-dimensional neural network model for processing, and obtaining an action extraction result of the video to be identified; determining, according to the action extraction result of the video to be identified, an action instance detection result of the video to be identified; inputting the video to be identified into a trained second three-dimensional neural network model for processing, and obtaining an action category discrimination result of the video to be identified; and determining an action category of the video to be identified according to the action instance detection result of the video to be identified and the action category discrimination result of the video to be identified. Combining different identification results obtained by using two three-dimensional neural network models may improve the identification efficiency of three-dimensional neural network models and reduce the calculation amount of a single three-dimensional neural network model.
Inventors:
Kikou
Wu Jialin
Yang Takeshi
Wang Valley
Wu Jialin
Yang Takeshi
Wang Valley
Application Number:
JP2020524869A
Publication Date:
August 18, 2021
Filing Date:
November 08, 2018
Export Citation:
Assignee:
Tsinghua University
International Classes:
G06T7/00; G06T7/20
Other References:
Rui Hou, Chen Chen, Mubarak Shah,Tube Convolutional Neural Network (T-CNN) for Action Detection in Videos,2017 IEEE International Conference on Computer Vision,IEEE,2017年,p5823-5832
Attorney, Agent or Firm:
try international patent corporation