Criminal emotion detection framework using convolutional neural network for public safety.

Area of Science:

• Computer Science
• Artificial Intelligence
• Forensic Science

Background:

• Modern societal changes and technological advancements necessitate proactive public safety measures.
• Understanding crime patterns and associated emotions is crucial for effective law enforcement and judicial processes.

Purpose of the Study:

• To propose a collaborative AI framework for detecting crime patterns and criminal emotions.
• To enhance judiciary decision-making through improved crime and emotion analysis.

Main Methods:

• Utilized two datasets: a crime dataset and an emotion dataset with 135 classes.
• Employed convolutional neural networks (CNNs) for image classification to detect crime and extract criminal faces.
• Applied various CNN architectures (LeNet-5, VGGNet, RestNet-50, basic CNN) for emotion detection from facial data.

Main Results:

• Achieved 92.45% accuracy in crime detection using CNN models.
• LeNet-5 demonstrated superior performance in criminal emotion detection, reaching 98.6% accuracy.
• Evaluated framework using metrics like accuracy, loss, precision-recall curve, and inference time.

Conclusions:

• The proposed CNN-based framework effectively detects crime and analyzes criminal emotions.
• This AI-driven approach shows significant potential for enhancing judicial decision-making and public safety applications.