e-ISSN: 2582-5208

International Research Journal of Modernization in Engineering Technology and Science

( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:05/Issue:12/December-2023 ImpactFactor-7.868 www.irjmets.com

FORENSIC FACE SKETCH CONSTRUCTION AND RECOGNITION

Akhilesh K A *1, Preetham N N*2, Sujay C L*3, Sumanth G G*4
*1,2,3,4Student Department of Computer Science & Engineering, KSSEM, Bangalore,
Karnataka, India.
ABSTRACT
In the realm of forensic science, the conventional practice of utilizing hand-drawn face
sketches for criminal identification has proven to be both time-intensive and somewhat
obsolete in our technologically advanced age. To tackle this issue, our project introduces an
independent software application that grants users the ability to craft composite facial
sketches of potential suspects effortlessly through an intuitive drag-and-drop interface.
Notably, this application harnesses the power of deep learning and cloud-based infrastructure
to facilitate swift and precise matching with law enforcement databases, effectively mitigating
the shortcomings of existing techniques and offering an adaptable, high-precision method for
criminal identification.
Keywords: Forensic science, drag-and-drop interface, composite facial sketches, cloud-based infrastructure
I. INTRODUCTION
The primary aim of this project is to develop a standalone application that empowers users to
construct composite face sketches of suspects efficiently, facilitating swift criminal
identification by matching police database. Develop an innovative and efficient forensic
application that revolutionizes the process of creating composite face sketches, matching them
with database records, and enhancing the accuracy and speed of criminal identification while
ensuring robust security measures. Additionally, the project aims to accelerate the matching
of these composite sketches with police databases, facilitating swift criminal identification.
Over the years, various techniques have been proposed to automate the process of identifying
suspects using hand-drawn face sketches. However, these methods have fallen short in
delivering precise results. Composite face sketch applications were introduced as an
alternative, but they too had limitations, such as a restricted facial features kit and a tendency
to produce cartoonish representations, making them less practical for law enforcement.
Motivated by the shortcomings of existing applications, we envisioned a solution that goes
beyond offering a standard set of facial features for users to select. Our innovative application
allows users to upload hand-drawn individual features, which are then seamlessly integrated
into the application's component set. This unique approach ensures that the created sketches
closely resemble hand-drawn counterparts, facilitating easier adaptation by law enforcement
departments. Notably, our application goes a step further by enabling law enforcement teams
to upload previously created hand-drawn sketches. Leveraging advanced deep learning
algorithms and cloud infrastructure, the platform enhances efficiency and accuracy in suspect
identification. The machine learning algorithm learns from the uploaded sketches and
database, providing users with suggested facial features that complement a single selected
feature. This not only streamlines the process but also significantly improves the overall
efficiency of the platform.

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[1]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:05/Issue:12/December-2023 ImpactFactor-7.868 www.irjmets.com
II. LITERATURE SURVEY
A. Composite sketch-based face recognition using ANN Classification:
The project provides a comprehensive overview of the significance of face detection and identification in
computer vision, particularly in the context of law enforcement and forensic applications. It highlights the
evolution of technology over the past 2 decades, emphasizing the challenges in developing computational
models for accurate face identification. The project discusses the various applications of face identification,
including security systems, credit card verification, and criminal investigations. It delineates the difference
between forensic sketches and composite sketches, with a focus on the prevalent use of software tools for
creating composite sketches in law enforcement. The project introduces key face recognition techniques such
as detection, extraction, and recognition, detailing the importance of moment functions and Local Binary
Patterns (LBP) in feature extraction. The proposed system is briefly outlined, involving the identification of
composite sketches through face region localization, facial component detection, feature extraction using
Orthogonal moments and MLBP, and classification using an artificial neural network (ANN) classifier.
Overall, the project serves as a literature survey encompassing the advancements and challenges in face
identification technology.
B. Forensic Face Photo-Sketch Recognition using a Deep Learning-based Architecture:
The project focuses on addressing challenges in automatically identifying subjects depicted in forensic
sketches, particularly when using real-world sketches and extended galleries resembling law enforcement
mug-shot databases. The approach involves leveraging deep learning for face photo-sketch recognition, a
domain where limited sketch images hinder the effective training of large networks. The project makes
several contributions to overcome these challenges, including the use of a pre-trained face photo recognition
model fine-tuned for sketch recognition through transfer learning. Additionally, a 3D morphable model is
employed to synthesize new images and expand the training dataset, preventing overfitting and enhancing
feature learning. Synthetic sketches are also introduced in the testing stage to improve performance, and the
proposed method is fused with a state-of-the-art algorithm for further enhancement. In the existing
literature, various methods for face photo-sketch recognition have been explored. Traditional approaches
often struggle with the modality gap and inaccuracies in sketches, leading to the development of specialized
algorithms for this specific task. Some reported methods have achieved high retrieval rates, but they often
rely on sketches closely resembling original photos, overlooking distortions caused by memory and
communication gaps. State-of-the-art methods in the literature often utilize hand-crafted features such as
the Scale-Invariant Feature Transform (SIFT) and Multiscale Local Binary Pattern (MLBP). However, these
features may not be optimal for inter-modality face recognition.
C. A Study on various state of the art of the Art Face Recognition System using Deep Learning
Techniques:
Techniques in face recognition and matching, emphasizing the significant improvements achieved by
training on extensive datasets. The use of deep learning models contributes to the accuracy enhancement of
face recognition systems. The review also encompasses a survey of various techniques employed to match
composite sketches to human images. This includes approaches such as component-based representation
and automatic composite sketch recognition. In the introduction section, the text highlights the two essential
patterns of face recognition systems: structural similarity and individual local differences of human faces. It
emphasizes the importance of feature extraction in the face detection process. The evolution of face
recognition is attributed to both technical challenges and its vast potential applications in video surveillance,
identity authorization, multimedia, security, law enforcement, and human-computer interaction. The final
portion of the introduction introduces facial recognition technology (FRT), noting its controversial nature
and significant impact on privacy. The text acknowledges the historical development of FRT since the 1960s
and its recent accessibility to law enforcement and private consumers. It also mentions the capacity of FRT
to replace traditional authentication methods like passwords, fingerprint data, and keys.

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[2]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:05/Issue:12/December-2023 ImpactFactor-7.868 www.irjmets.com
D. Face recognition accuracy of forensic examiners, super recognizers, and face recognition
algorithms:
The study involves a comprehensive comparison of face identification by forensic facial examiners, facial
reviewers, superrecognizers, fingerprint examiners, and students. Additionally, it explores the performance
of deep convolutional neural networks (DCNNs) developed between 2015 and 2017. The results highlight
the variability in individual performance and the increasing accuracy of algorithms over time. Forensic facial
examiners, facial reviewers, and superrecognizers were found to be more accurate than fingerprint
examiners and students in a challenging face identification test. The individual performance varied widely
among human participants. Four deep convolutional neural networks (DCNNs) developed between 2015 and
2017 were tested, and their accuracy in identifying faces was within the range of human accuracy. The most
recent DCNN outperformed the median accuracy of forensic facial examiners. The text acknowledges the
scarcity of research on the accuracy of forensic facial examiners relative to people without training and the
lack of knowledge about their accuracy relative to computer-based face recognition systems. It refers to the
need for independent and objective scientific research in this domain. The literature survey touches upon
the evolution of face recognition technology over the last decade, mentioning the gradual closing of the
performance gap between human and machine-based systems on challenging face recognition tasks.
E. Face Photo - Sketch Construction and Recognition:
In recent developments within the realms of computer vision and biometrics, the creation and recognition
of face sketches have emerged as pivotal subjects. Face sketches, whether hand-drawn or digitally generated,
focus solely on capturing facial features, omitting nuances like skin tone. These sketches play a crucial role
in forensic investigations and law enforcement for identifying suspects or missing persons. Current research
delves into algorithms that autonomously generate face sketches from images and employ computer vision
and deep learning models for precise facial trait extraction and comparison against a database of recognized
faces. This literature explores the latest techniques in face sketch creation and recognition, discussing
applications, challenges, and potential future advancements.
III. METHODOLOGY
A. Create User-Friendly Application:
Conduct user surveys and interviews to understand user preferences and challenges. Design an intuitive
and easy-to-navigate interface based on the feedback received. Implement prototype testing and gather
user feedback for iterative improvements.
B. Data Collection:
Collaborate with law enforcement agencies to collect existing sketch databases. Implement a secure data
transfer mechanism, ensuring privacy and adherence to legal regulations. Develop features for users to
upload hand-drawn sketches, employing image processing techniques for data standardization.
C. Implement Advanced Facial Recognition:
Employ deep learning algorithms for facial feature extraction from both hand-drawn sketches and composite
sketches. Train the model on a diverse dataset to enhance recognition accuracy. Implement a two-step
verification process for robust identity confirmation.
D. Utilize Cloud Infrastructure:
Select a reliable cloud service provider and set up a secure cloud infrastructure. Implement data encryption
and access controls to ensure the confidentiality of sensitive information. Optimize the application for
seamless integration with cloud services, enabling efficient storage and retrieval of facial recognition data.
Regularly update and maintain the cloud infrastructure for scalability and security.
IV. RESULTS AND DISCUSSION
The 'Forensic Face Sketch Construction and Recognition' project excels in security, privacy, and accuracy.
Security measures, such as MAC Address and IP Address matching, coupled with an effective OTP system,
ensure stringent access controls. During testing, the platform achieves an impressive average accuracy rate
exceeding 90% and a confidence level of 100%, outperforming related studies. Unique features contribute
www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science
[3]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:05/Issue:12/December-2023 ImpactFactor-7.868 www.irjmets.com
to enhanced security and accuracy, setting the project apart. In summary, the project offers a robust and
reliable solution, demonstrating efficacy in real-world scenarios for law enforcement agencies in forensic
face sketch construction and recognition.
V. CONCLUSION
In conclusion, the 'Forensic Face Sketch Construction and Recognition' project stands out for its robust
security measures, achieving a high level of accuracy above 90% with a confidence level of 100%. The
unique features implemented contribute to the platform's effectiveness, providing law enforcement
agencies with a reliable solution for forensic face sketch construction and recognition in real-world
scenarios.
VI. REFERENCES
[1] Forensic Face Sketch Construction and Recognition Asst. Prof. Abhijit Patil [1] , Akash Sahu [2] , Jyoti
Sah [3] , Supriya Sarvade [4] , Saurabh Vadekar [5]-2020
[2] Composite Sketch Based Face Recognition Using ANN Classification Shivaleela Patil, Dr.Shibhangi D C -
2020
[3]Forensic Face Photo-Sketch Recognition using a Deep Learning-based Architecture Christian Galea,
Student Member, IEEE, and Reuben A. Farrugia, Member, IEEE-2017
[4]He, D.C. and Wang, L., 1990. Texture unit, texture spectrum, and texture analysis. IEEE transactions on
Geoscience and Remote Sensing, 28(4)-2019
[5]Face recognition accuracy of forensic examiners, superrecognizers, and face recognition algorithms P.
Jonathon Phillipsa,,, Amy N. Yatesa, Ying Hub -2017

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[4]