Project Synopsis Forensic Face Sketch Construction And Recognition

Project Synopsis
PROJECT TITLE
Forensic Face Sketch Construction and Recognition using ML

PROJECT LOCATION
APS College of Engineering, CSE Department

INTRENAL GUIDE
Prof. Soumya George

TEAM MEMBERS
Dishan D A (1AP20CS009)
-9570799171

S Naveen (1AP20CS026)
-9353458267
Vinay Kumar (1AP20CS032)
-8935837289

PROJECT DURATION
Start Date: 1st December 2023
End Date: 25th May 2024

REQUIREMENTS
Software Requirements:

• Operating System: Windows 8.0 and above.

Hardware Requirements:

• Processor: intel i3 or equivalent

• RAM: 6 GB or more
• Storage: 128 GB or more
• Screen Resolution: 720 x 1280 pixels or higher
• Sensors: Accelerometer, Gyroscope
• Connectivity: Wi-Fi, Mobile Data
Project Synopsis Forensic Face Sketch Construction And Recognition

PROBLEM STATEMENT
Develop a machine learning-based system for Forensic Face Sketch Construction and Recognition. Challenges
include subjective eyewitness descriptions, artistic style variations, and real-time optimization for sketch
construction. The Face Recognition component must handle lighting variations, image quality issues, and
robustly identify suspects in law enforcement databases. Ethical considerations involve bias mitigation, privacy
protection, and responsible technology use. The project aims to create an advanced, accurate, and ethical solution
for criminal investigations, addressing challenges through comprehensive metrics and evaluations.

EXISTING SYSTEM
One notable system is "FACES" (Faces, Anticipating, and Criminals, Evaluation, and Systems). FACES utilizes
deep learning techniques to generate facial sketches from descriptions and then employs facial recognition
algorithms for suspect identification. It addresses challenges such as variable artistic interpretations and real-
world applicability. However, given the rapidly evolving nature of technology, new systems may have emerged
or existing ones may have undergone updates.

PROPOSED SYSTEM
Forensic Face Sketch Construction and Recognition is a proposed system integrating advanced machine learning
for precise forensic face sketch construction and recognition. Key features include a DeepSketch Generator for real-
time, detailed sketches from eyewitness descriptions with context-aware features. The system includes a Sketch
Matching Module for accurate facial recognition against dynamic law enforcement databases, addressing lighting and
disguise challenges. Ethical considerations involve bias mitigation and privacy protection, adhering to guidelines.
Performance evaluation incorporates comprehensive metrics and real-world simulations. The user-friendly interface
streamlines interactions for law enforcement. Forensic Face Sketch Construction and Recognition aims for accuracy,
ethical integrity, and practical usability in criminal investigations.

Objective:
1. Develop a DeepSketch Generator using deep learning to translate eyewitness descriptions into detailed facial
sketches in real-time.
2. Implement context-aware features to enhance the realism of generated sketches, considering variables such as
age, emotions, and facial expressions.
3. Create a Sketch Matching Module that integrates facial recognition algorithms, tailored for matching forensic
sketches against dynamic law enforcement databases.
4. Address ethical considerations by incorporating bias mitigation techniques in both training data and
algorithms, ensuring fairness across demographic groups.
5. Design a user-friendly interface for Forensic Face Sketch Application, streamlining interactions for law
enforcement personnel, facilitating easy input of eyewitness descriptions, and seamless integration with
existing workflows.

METHODOLOGY
1. Data Collection:
• Gather a diverse dataset comprising facial images, corresponding sketches, and contextual information.
• Curate eyewitness descriptions to create a comprehensive training dataset.
2. DeepSketch Generator:
• Implement a deep learning architecture (e.g., Convolutional Neural Network) for the DeepSketch
Generator.
• Train the model using the curated dataset, optimizing for accuracy and real-time performance.
• Incorporate context-aware features by including additional information in the training data and adjusting
the model architecture.
Project Synopsis Forensic Face Sketch Construction And Recognition

3. Sketch Matching Module :

• Design a facial recognition module tailored for matching forensic sketches against law enforcement
databases.
• Integrate state-of-the-art facial recognition algorithms, considering variations in lighting, angles, and
disguises.
• Train and fine-tune the module using a diverse set of facial images and sketches.
4. Ethical Consideration:
• Implement bias mitigation techniques such as adversarial training to reduce biases in the training data
and model.
• Incorporate privacy-preserving measures, including anonymization of non-criminal individuals in the
database.
5. Performance Evaluation:
• Define comprehensive metrics (accuracy, precision, recall) to evaluate the performance of the Deep
Sketch Generator and Sketch Matching Module.
• Conduct simulations that replicate real-world forensic scenarios to assess the system's practical
effectiveness.
6. User Interface Development:
• Design an intuitive and user-friendly interface for law enforcement personnel.
• Enable easy input of eyewitness descriptions and retrieval of generated sketches.
• Ensure seamless integration with existing law enforcement workflows and databases.
7. Iterative Testing and Optimization:
• Conduct iterative testing to identify and address system weaknesses.
• Optimize the system based on feedback from users and ongoing evaluations.
• Incorporate continuous updates to adapt to emerging challenges and advancements in machine learning.
8. Deployment and Integration:
• Deploy Forensic application in a controlled environment for initial testing and refinement.
• Integrate the system into existing law enforcement infrastructures, ensuring compatibility and minimal
disruption.
9. Documentation and Training:
• Provide comprehensive documentation for users, including guidelines on system operation and ethical
considerations.

Conclusion
Forensic Face Sketch Construction and Recognition presents a comprehensive and advanced solution for forensic
face sketch construction and recognition, leveraging state-of-the-art machine learning techniques. The proposed
system addresses key challenges in accuracy, ethical considerations, and usability for law enforcement. The
DeepSketch Generator, with its real-time capabilities and context-aware features, enhances the fidelity of
generated sketches. The Sketch Matching Module is tailored for dynamic recognition against diverse law
enforcement databases, considering real-world challenges. Ethical considerations, including bias mitigation and
privacy protection, are embedded in the system design. The user-friendly interface ensures seamless integration
with existing workflows. Through iterative testing, optimization, and a structured deployment process, Forensic
Face Sketch Construction and Recognition aims to be a reliable, ethical, and practical tool for enhancing criminal
investigations in the modern era.
Project Synopsis Forensic Face Sketch Construction And Recognition