A Decision Support System for Face Sketch Synthesis Using Deep Learning and Artificial Intelligence.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Cybersecurity

Background:

• The increasing crime rate necessitates advanced investigative tools.
• Internet of Things (IoT) technologies offer potential for suspect identification.
• Existing methods for face sketch synthesis using IoT and deep learning are limited.

Purpose of the Study:

• To develop an improved deep learning architecture for face sketch synthesis.
• To enhance suspect identification capabilities for law enforcement.

Main Methods:

• A novel neural network architecture, Spiral-Net (a U-Net variant), was developed for face sketch synthesis.
• Spiral-Net integrates with a pre-trained Vgg-19 network (feature extractor F) and a custom CNN (discriminator D).
• The system identifies top sketch matches, formulates feature maps, and uses loss functions to iteratively update the networks.

Main Results:

• The Spiral-Net model demonstrated a 5% improvement in face recognition accuracy compared to the state-of-the-art.
• Performance was evaluated using datasets like CUFS, CUFSF, and IIT photo-sketch.
• Results were validated against established face recognition schemes (NLDA and OpenBR).

Conclusions:

• The proposed Spiral-Net architecture offers a significant advancement in face sketch synthesis.
• This technology has strong potential for application in law enforcement and cybersecurity.
• Further development could enhance its utility in combating cybercrime.