Transfer learning for the classification of video-recorded crowd movements
- 1 September 2019
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 271-276
Abstract
The automatic recognition of a crowd movement captured by a CCTV camera can be of considerable help to security forces whose mission is to ensure the safety of people on the public area. In this context, we propose to fine-tune a model from the TwoStream Inflated 3D architecture, pre-trained on the ImageNet and the Kinetics source datasets, to classify video sequences of crowd movements from the Crowd-11 target dataset. The evaluation of our model demonstrates its superiority over the state-of-the-art in terms of classification accuracy.Keywords
This publication has 25 references indexed in Scilit:
- Learning Spatiotemporal Features with 3D Convolutional Networks2015
- Socially Constrained Structural Learning for Groups Detection in CrowdIEEE Transactions on Pattern Analysis and Machine Intelligence, 2015
- Crowded Scene Analysis: A SurveyIEEE Transactions on Circuits and Systems for Video Technology, 2014
- Action Recognition with Improved Trajectories2013
- A Literature Review on Video Analytics of Crowded Scenes2013
- Loveparade 2010: Automatic video analysis of a crowd disasterComputer Vision and Image Understanding, 2012
- Crowd behavior analysis under cameras network fusion using probabilistic methods2010
- A Survey on Transfer LearningIEEE Transactions on Knowledge and Data Engineering, 2009
- Crowd analysis: a surveyMachine Vision and Applications, 2008
- A Duality Based Approach for Realtime TV-L 1 Optical Flow2007