Optimized Convolutional-Recurrent Architecture for Detecting Diverse Crimes in Real-Time

Rasool, Zohaib (2024) Optimized Convolutional-Recurrent Architecture for Detecting Diverse Crimes in Real-Time. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

This research proposes a new solution to real-time crime detection by expanding the Convolutional Recurrent Auto Encoder (CR-AE) model to detect 12 different types of crimes and typical scenes based on video surveillance data. This work utilizes an incremental model development approach where Conv3D, ConvLSTM2D, and Conv3DTranspose are used to capture spatio-temporal features. Out of the four versions of the model developed the best performing model was Version 2 (V2) when the data was split into 80-10-10 data split. The proposed model testing accuracy was 84.34% with AUC and F1-score being 0.97 and 0.83 respectively, suggesting it would be useful for feature extraction and computational requirements. V2 showed an intermediate depth which allowed it to generalize well across different crime scenes as it outperformed models with higher depth like V4 and V5 that seemed to over-fit. The trade-off relationship found in this work between model complexity and available data is critical if the performance of models is to be maximized. There is an evident efficiency of the suggested system, yet, its weaknesses include the usage of video data only and occasional classification errors. Potential future work can include the addition of audio features, more data augmentation and the use of attention mechanisms to increase resilience and architecture flexibility. This research provides a number of improvements over previous work in automated crime detection and presents a solid basis for the application of intelligent surveillance in more realistic environments.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Milosavljevic, Vladimir UNSPECIFIED
Subjects:	H Social Sciences > HV Social pathology. Social and public welfare > Criminology > Crimes and Offences Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	04 Sep 2025 13:16
Last Modified:	04 Sep 2025 13:16
URI:	https://norma.ncirl.ie/id/eprint/8788

Actions (login required)

View Item