Crime type identification from video footage based on 3D convolutional neural network

Loading...
Thumbnail Image

Date

2022

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

In this thesis, we investigate the problem of crime classification given video footage from CCTV cameras. To do this, we will use 3D convolutional neural networks that first intended to be used for recognizing various actions. The unique feature of 3D convolutional neural networks lies in the fact that they capture both spatial and temporal dimensions by using 3D convolution, thus processing the motion data and its outlines. The system discussed in this thesis produces distinct channels for incoming data based on the input sketches. The final part includes representation that aggregates data from all channels. It is shown that the discussed architecture results a high accuracy in identifying the type of criminal offence shown in the underlying video footage.

Description

Keywords

crime, CCTV cameras, 3D convolutional neural networks, video footage

Citation

Collections