Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
INFORMATION PROCESSING METHOD, INFORMATION PROCESSING DEVICE, AND PROGRAM
Document Type and Number:
WIPO Patent Application WO/2018/088170
Kind Code:
A1
Abstract:
Provided is an information processing method, comprising: an input step (S1) of inputting a video into a neural network; a processing step (S2) of carrying out a convolution process upon the present frame which is included in the video and computing a present characteristic map which is a characteristic map of the present, combining the present characteristic map with a prior characteristic map which is obtained by carrying out a convolution process upon a prior frame which is included in the video, using the combined prior characteristic map and present characteristic map to infer a candidate object region, and using the combined prior characteristic map and present characteristic map together with the inferred candidate object region to infer positions and identification information of one or more objects which appear in the present frame; and an output step (S3) of outputting as an object detection result the positions and identification information which have been inferred in the processing step (S2) of the one or more objects which appear in the present frame.

Inventors:
SENAY GREGORY
TSUKIZAWA SOTARO
Application Number:
PCT/JP2017/037937
Publication Date:
May 17, 2018
Filing Date:
October 20, 2017
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
PANASONIC IP MAN CO LTD (JP)
International Classes:
G06T7/00; G06V10/764
Foreign References:
JP2004171159A2004-06-17
Other References:
HUKUI, HIROSHI ET AL.: "Random Dropout and Ensemble Inference Networks for Pedestrian Detection and Traffic Sign Recognition", TRANSACTIONS OF INFORMATION PROCESSING SOCIETY OF JAPAN, vol. 57, no. 3, 15 March 2016 (2016-03-15), pages 910 - 921, XP009515166, ISSN: 1882-7764
SHAOQING REN: "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks", INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV, 2015, pages 1440 - 1448
LANDOLA F. N.: "SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and< MB model size", ARXIV PREPRINT, ARXIV: 1602.07360, 2016
KAIMING HE: "Deep Residual Learning for Image Recognition", THE IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR, 2016, pages 770 - 778, XP055536240, DOI: doi:10.1109/CVPR.2016.90
ALEX KRIZHEVSKY: "ImageNet Classification with Deep Convolutional Neural Networks", PART OF: ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS, vol. 25, 2012
Attorney, Agent or Firm:
KAMATA Kenji et al. (JP)
Download PDF:



 
Previous Patent: HEAT EXCHANGER

Next Patent: REINFORCING FILM WITH SEPARATOR