The document proposes a DeepCNN-based criminal identification system called DeepFace to help law enforcement identify suspects more accurately and efficiently. Some key points:
- Current manual systems are prone to human error while existing automated systems have accuracy issues dealing with variations in faces.
- DeepFace aims to develop an accurate and reliable DeepCNN model to identify criminals from facial images by standardizing preprocessing, using RPN for face detection, GLCM for feature extraction, and DCNN for classification.
- The proposed system is expected to provide improved identification accuracy, be scalable to large databases, and automate the identification process to save law enforcement resources compared to existing solutions.
Essential UI/UX Design Principles: A Comprehensive Guide
Criminal Identification - Full PPT (1) (1).pptx
1. DeepFace: Deep Learning Model based
Criminal Identifications System for Law
Enforcement Department
2. Problem Identified
• A criminal identification system is a technology or process used to identify
individuals who have been involved in criminal activities.
• major problem with manual systems of criminal face identification for law
enforcement departments is the potential for human error and subjective
judgments.
• automated systems using machine learning algorithms for criminal face
identification by law enforcement departments is the potential for
inaccuracies and biases in the algorithms used to analyze and match faces.
• criminal identification systems can be a valuable tool for law enforcement
agencies in identifying suspects and solving crimes, but there are concerns
about privacy, accuracy, and bias that need to be addressed.
3. Objective
The aim of developing a criminal face identification system using
DeepCNN for law enforcement department is to provide an accurate,
reliable, and efficient tool for identifying suspects involved in criminal
activities.
• To create a comprehensive database of facial images for training and testing the
DeepCNN model that includes a diverse range of individuals.
• To pre-process facial images to ensure that they are standardized for lighting,
orientation, and other factors that may affect the accuracy of the system.
• To develop and optimize a DeepCNN model that can accurately identify
individuals based on facial features and patterns, with a focus on achieving high
accuracy and speed.
4. Abstract
• The current criminal face identification system for law enforcement departments is
largely manual and relies on human judgment, which is time-consuming, error-prone, and
subject to biases.
• There is a need for an automated system that can quickly and accurately identify criminal
suspects based on facial recognition.
• Existing facial recognition systems often suffer from low accuracy and speed, particularly
when faced with variations in lighting, pose, and facial expressions.
• This can result in false positives or negatives, leading to wrongful accusations or missed
opportunities to apprehend criminals.
• DeepFace is to develop an accurate and efficient criminal face identification system using
DeepCNN that can overcome the limitations of current systems, enabling law
enforcement agencies to quickly and accurately identify criminal suspects and enhance
public safety.
5. Introduction
• Criminal face identification is a crucial task for law enforcement
departments, enabling the identification and apprehension of suspects
involved in criminal activities.
• DeepCNN is a type of deep learning algorithm that has shown great promise
in achieving high accuracy and speed in facial recognition tasks.
• DeepFace is a criminal face identification system that utilizes DeepCNN to
accurately and efficiently identify criminal suspects for law enforcement
departments.
• In this project, we aim to develop a robust and efficient criminal face
identification system using DeepCNN that complies with ethical and legal
requirements, including data privacy and non-discrimination laws.
6. Existing System
• The current manual system for criminal face identification used by law
enforcement departments involves a human operator visually comparing a
suspect's face to a database of known criminal images or sketches.
• Several automated systems using machine learning algorithms have been
developed for criminal face identification in recent years.
• Some of the commonly used algorithms include K-Nearest Neighbors (KNN),
Support Vector Machines (SVM), Decision Tree, and Logistic Regression.
7. Disadvantages
• Manual process is time-consuming, and it can take a long time to identify a suspect
accurately.
• The manual system is prone to errors, particularly when faced with variations in lighting,
pose, and facial expressions.
• The manual system is costly, as it requires a large workforce to manually compare the images,
and it can be difficult to maintain the accuracy of the identification process over time.
• Existing algorithms may have limited accuracy when faced with variations in lighting, pose,
and facial expressions.
• This can lead to misidentification of criminal suspects, which can have severe consequences.
• These algorithms can be complex and challenging to understand, which can make it difficult
for law enforcement officials to evaluate the accuracy and reliability of the system.
8. Proposed System
The proposed system for DeepFace: Criminal Face Identifications for Law Enforcement
Department using DeepCNN consists of several stages, including:
Preprocessing: The input images are preprocessed to remove any noise, standardize
lighting conditions, and resize the images to a fixed size.
Face detection using RPN: A Region Proposal Network (RPN) is used to identify potential
faces in the preprocessed images.
Feature Extraction using GLCM: The Gray-Level Co-occurrence Matrix (GLCM) is used
to extract features from the potential faces detected by the RPN.
Classification and Identification using DCNN: A Deep Convolutional Neural Network
(DCNN) is used for classification and identification. Criminal with name and Details or
non-criminal and identify the suspect based on matching the face to a database of known
criminals.
9. Advantages
Improved accuracy: By using advanced preprocessing techniques, state-of-the-art
face detection using RPN, feature extraction using GLCM, and deep learning-based
classification and identification using DCNN, the proposed system can provide
improved accuracy in identifying criminal faces.
The proposed system is scalable and can handle large databases of criminal images,
making it suitable for use in real-world scenarios.
Automated: The proposed system is fully automated, which can save time and
resources for law enforcement departments.
The proposed system can perform criminal face identification quickly and
efficiently, allowing law enforcement departments to act on the information they
receive more rapidly.
10. System Specifications
Server Side : Python 3.7.4(64-bit) or (32-bit)
Client Side : HTML, CSS, Bootstrap
IDE : Flask 1.1.1
Back end : MySQL 5.
Server : Wampserver 2i
Packages : TensorFlow, Keras, Pandas, SiKit Learn
11. System Architecture
Face Enrollment and Updating
Capture Face
Convert into Frames
Preprocessing
RPN Face Detection
Feature Extraction
CNN Classification
LED Server
Prediction
Predict Criminal Face
Criminal Face
Update Face Details
Alert
Criminal Biodata
Crime History
Input Image or
Live Video
12. Modules
1. Criminal Identification Dashboard
2. End User Interface
3. Criminal Face Classification
3.1. Upload Datasets
3.2. Pre-processing
3.3. Face Detection
3.3. Feature Extraction
3.4. CNN Classification
4. Criminal Face Identification
4.1. Upload Crime Photo or Live Video
4.2. Identify Criminal
5. Alert
6. Performance Analysis
13. 1. Criminal Identification SystemWeb App
The Criminal Identification System Web App is a web-based application that is
designed to identify and track criminals using facial recognition technology.
The application is built using Python Flask, a popular web framework, and Tensor
Flow, an open-source machine learning library.
The Criminal Identification System Web App is designed to be user-friendly and
efficient, providing law enforcement agencies with a powerful tool to track and
apprehend criminals.
The dashboard can be customized to meet the specific needs of each law
enforcement agency, providing a powerful tool to track and apprehend criminals.
14. 2. End User Dashboard
The End User Dashboard module for DeepFace: Criminal Face Identifications for
Law Enforcement Department using DeepCNN is designed to provide a user-
friendly interface for end users to perform facial recognition searches against the
database of criminal records.
2.1. Web Admin: The web admin is responsible for managing the overall system,
including user management, data management, and system configuration. The
admin would have access to the system's backend, where they can configure the
system's settings and manage the system's database.
2.2. Law Enforcement Officers: Law enforcement officers are the primary users of
the system, who would use the system to identify suspects and track down criminals.
They would have access to the system's frontend, where they can search the system's
database, upload images, and receive identification results.
15. 3. Criminal Face Classification
3.1. Data collection: Images of criminal faces are collected from
various sources, such as surveillance cameras, criminal databases, and
social media platforms.
3.2. Data preprocessing: The first step is to preprocess the input
images. The RGB images are converted to grayscale, resized to a
standard size, denoised, and binarized to enhance the features.
3.3. Region Proposal Network (RPN) face detection: The next step is
to detect the face regions in the preprocessed images using the RPN face
detection algorithm. The RPN is a deep learning-based algorithm that
proposes regions in an image that are likely to contain a face.
16. 3. Criminal Face Classification
3.4. Gray-Level Co-occurrence Matrix (GLCM) feature extraction:
Once the face regions are detected, the next step is to extract features
from the face images using the GLCM technique. GLCM is a method of
feature extraction that quantifies the texture properties of an image by
calculating the co-occurrence matrix of gray levels.
3.5. Convolutional Neural Network (CNN) classification: The final
step is to classify the face images into their respective categories using a
CNN. The CNN is a deep learning-based algorithm that can learn the
features of the face images from the GLCM features extracted in the
previous step.
17. 4. Criminal Face Identification
Criminal Face Identification is the process of matching an input image of a person's
face with a known identity in a database. In other words, it involves identifying a person
based on their face in an image.
4.1. Input Image or Live Video
The input image or live video is first processed through various stages such as
preprocessing, face detection, and feature extraction. The resulting features are then
compared with the features of known identities in a database using similarity measures
such as Euclidean distance or cosine similarity.
4.2. Identification
If a match is found, the identification of the person in the input image is returned. If no
match is found, the system may return a list of possible matches or indicate that the
input image does not correspond to any known identity in the database.
This process is typically used in law enforcement and security applications, where it is
important to quickly and accurately identify individuals from images or video footage.
18. 5. Alert
An alert to law enforcement officers for criminal face identification can be triggered in the following
situations:
• If the identified person is a wanted criminal or has a criminal record, an alert should be
immediately sent to the officer in charge of the case.
• If the identified person is a missing person or a victim of a crime, an alert should be sent to the
officer in charge of the case.
• If there is a potential match between the identified person and a suspect in an ongoing
investigation, an alert should be sent to the officer in charge of the investigation.
• If the identified person is a known associate of a criminal, an alert should be sent to the officer in
charge of the case or the investigation.
• If the identified person is on a watchlist, an alert should be sent to the officer in charge of the
watchlist.
19. 6. Performance Analysis
Use evaluation metrics such as accuracy, precision, recall, and F1 score to measure
the performance of the DCNN model. These metrics will help in understanding how
well the model is performing in correctly identifying criminal faces.
22. Criminal Face
Identification
System
Web Admin
Police
Criminal Face
Predictor
engine
Criminal Face
DB
Load & Explore Dataset
Login
Login
Upload Criminal
Image or Fed
Live Video
Preprocessing
RPN Face Detection
GLCM Feature Extraction
Train & Build the Model
Identify Criminal Face
Alert and Give Info
Upload Criminal Face
Dataset
DCNN Classification
23. Level 2
Criminal Face
Identification
System
Web Admin
Police
Criminal Face
Predictor
engine
Criminal Face
DB
Load & Explore Dataset
Login
Login
Upload Criminal
Image or Fed
Live Video
Preprocessing
RPN Face Detection
GLCM Feature Extraction
Train & Build the Model
Identify Criminal Face
Alert and Give Info
Upload Criminal Face
Dataset
DCNN Classification
Alert
the
Police,
about
the
Identified
Criminal
Face
and
Product
Bio
Crime
History
24. Conclusion
• In conclusion, the proposed system for DeepFace: Criminal Face
Identifications for Law Enforcement Department using DeepCNN
offers significant improvements over existing manual and automated
systems for identifying criminal faces.
• This system can help law enforcement departments to identify and
apprehend suspects more effectively, and ultimately protect
communities from criminal activities.
• Overall, the proposed system using DeepCNN can significantly
enhance the capabilities of law enforcement departments in identifying
criminal faces, and has the potential to contribute to the reduction of
crime rates and the improvement of public safety.
25. Future Enhancement
• Integration with surveillance systems: The system can be integrated
with existing surveillance systems, allowing law enforcement to
monitor public areas and identify criminals in real-time.
• Deployment in other fields: The technology used in the proposed
system can be applied in other fields such as healthcare, finance, and
retail, for applications such as facial recognition-based access control
and customer identification.
• Multi-modal identification: The system can be extended to perform
multi-modal identification by integrating other biometric data such as
fingerprints, voice, and iris scans to enhance identification accuracy.
26. References
1. Zhang, X., et al. "Joint Face Detection and Alignment Using Multitask Cascaded Convolutional Networks."
IEEE Signal Processing Letters, vol. 23, no. 10, 2016, pp. 1499-1503.
2. Yang, J., et al. "GLCM Based Feature Extraction for Face Recognition." 2018 IEEE International Conference
on Applied System Innovation (ICASI), 2018, pp. 296-298.
3. Zhang, K., et al. "Deep Residual Learning for Image Recognition." Proceedings of the IEEE Conference on
Computer Vision and Pattern Recognition, 2016, pp. 770-778.
4. Liu, W., et al. "Rethinking the Smaller-Norm-Less-Informative Assumption in Channel Pruning of
Convolution Layers." arXiv preprint arXiv:1802.00124, 2018.
5. Wang, Y., et al. "Learning Face Age Progression: A Pyramid Architecture of GANs." Proceedings of the IEEE
Conference on Computer Vision and Pattern Recognition, 2018, pp. 31-39.