VIOLENCE DETECTION USING DEEP

LEARNING
NAME OF THE STUDENTS
1. Pratik Anand (23013440101)
2. Praveen Hembrom (23013440102)
3. Praveen Kumar Mishra (23013440103)
4. Priyanshu Kumar (23013440105)
5. Priyanshu Kumar (23013440106)

DEPARTMENT OF COMPUTER SCIENCE &

ENGINEERING
R.V.S. COLLEGE OF ENGINEERING AND TECHNOLOGY
JHARKHAND UNIVERSITY OF TECHNOLOGY

2023-2027
 A
Mini Project Report
On

(VIOLENCE DETECTION USING DEEP LEARNING)

Submitted in partial fulfillment of the requirements
for
3rd Semester
Bachelor of Technology in
Computer Science and Engineering

by
1. (Pratik Anand) University Roll No- (230131440101)
2. (Praveen Hembrom) University Roll No-
(230131440102)
3. (Praveen Kumar Mishra) University Roll No-
(230131440103)
4. (Priyanshu Kumar) University Roll No-
(230131440105)
5. (Priyanshu Kumar) University Roll No-
(230131440106)

Under the guidance of

Prof. Nushrat Praveen
Dept. of Computer Science and Engineering

Department Of Computer Science and Engineering

R.V.S. College of Engineering and Technology
Jamshedpur – 831012
2023 – 2027
 Department of Computer Science and Engineering
CERTIFICATE
This is to certify that the mini project work entitled

(VIOLENCE DETECTION USING DEEP LEARNING ) is done

by (Pratik Anand, Praveen Hembrom, Praveen Kumaar

Mishra, Priyanshu Kumar, Priyanshu Kumar), Regd. No:-

(23013440101, 23013440102, 23013440103,

23013440105, 23013440106) in partial fulfilment of the

requirements for the 3rd Semester Sessional Examination of

Bachelor of Technology in Computer Science &

Engineering during the academic year 2023 -27. This work

is submitted to the department as a part of evaluation of 3 rd

Semester Mini Project.

Prof. Nushrat Praveen

Department of CSE
Prof. Jeevan Kumar
Mini Project Guide H.O.D, CSE
 Signature of External

Date: 25/02/2025
Place: RVSCET, Jamshedpur
 DECLARATION
We declare that this written submission represents our ideas in our own
words and where others ideas or words have been included. We have
adequately cited and referenced the original sources. We also declare that
we have adhered to all principles of academic honesty and integrity and
have not misrepresented or fabricated or falsified any idea/data/fact/source
in our submission. We understand that any violation of the above will cause
for disciplinary action by the Institute and can also evoke penal action from
the sources which have thus not been properly cited or from whom proper
permission has not been taken when needed.

(Signature)
1. Pratik Anand

(Signature)
2. Praveen Hembrom

(Signature)
3. Praveen Kumar Mishra

(Signature)
4. Priyanshu Kumar

(Signature)
5. Priyanshu Kumar
 ACKNOWLEDGEMENT
We avail this opportunity to express our profound sense of
gratitude to Prof. Nushrat Praveen for her constant
supervision, guidance and encouragement right from the
beginning till the completion of the mini project.

We wish to thank all the faculty members and supporting

staffs of Dept. of Computer Science and Engineering for their
valuable suggestions, timely advice and encouragement.

(Signature)
1. Pratik Anand

(Signature)
2. Praveen Hembrom

(Signature)
3. Praveen Kumar Mishra

(Signature)
4. Priyanshu Kumar

(Signature)
Priyanshu Kumar
 Content
1. Introduction…………………………Pg. No. 01
 Purpose……………………………..
 Project Scope…………………………
 Product Features
2. System Analysis
 Software Requirements
 Hardware Requirements
 User (Questionnaires)
3. System Design & Specifications
High Level Design (HLD)
 Project Model
 Structure Chart
 DFD
 E-R Diagram
 UML ( Use Case Diagram)
Low Level Design (LLD)
 Screen-Shot Diagram
4. Coding
 Sample Code
5. Testing
 Unit Testing
 Integration Testing
6. Conclusion & Limitations.
7. Reference/Bibliography
 Introduction
Introduction to Violence Detection Using Deep Learning

In recent years, the detection of violence in various environments has become an

increasingly important concern, especially for enhancing public safety and
providing timely interventions. With the rise in the use of surveillance systems
across public spaces, the need to automatically detect violent behaviours and
respond accordingly has grown significantly. This project aims to leverage the
power of deep learning techniques to create a robust and efficient violence
detection system that can recognize violent actions from video footage in real-time.

Purpose:
The primary purpose of this project is to design and develop a deep learning-based
system capable of identifying violent behaviours such as physical altercations,
aggression, or other dangerous activities. This technology aims to enhance safety
protocols in public areas such as shopping malls, schools, and transportation
systems, by automatically alerting authorities when violence is detected, enabling
quicker responses to prevent escalation.

Project Scope:
The scope of this project involves the following key aspects:

1. Dataset Collection and Preprocessing: Gathering and preprocessing large

datasets of labelled video data that includes violent and non-violent actions.
2. Model Development: Designing deep learning models, particularly
convolutional neural networks (CNNs) and recurrent neural networks
(RNNs), to identify patterns in videos that correspond to violent actions.
3. Real-Time Detection: Building a system capable of processing video
streams in real-time and detecting violent incidents as they happen.
4. Performance Evaluation: Testing the developed system's accuracy,
precision, recall, and real-time response capabilities under various
conditions and environments.
Product Features:

 Real-Time Monitoring: The system will provide continuous surveillance,

with instant detection and alert systems triggered by violent activities.
 Automated Alerts: Upon detecting violent behaviour, the system will
automatically notify security personnel or local authorities, allowing for
faster intervention.
 High Accuracy: By training deep learning models with large, diverse
datasets, the system will achieve high accuracy and minimize false positives
or negatives.
 Adaptability: The system can be integrated with existing surveillance
infrastructure, adaptable for use in a variety of settings, including urban
areas, corporate spaces, or public events.
 Scalability: The solution will be scalable to accommodate varying numbers
of cameras and diverse environments without compromising performance.
 System Analysis
Software Requirements

 Programming Languages: Python, TensorFlow/Keras, OpenCV

 Libraries:
o TensorFlow / PyTorch (for deep learning model development)
o OpenCV (for video processing)
o NumPy, Matplotlib (for data manipulation and visualization)
o Scikit-learn (for model evaluation and metrics)
 Tools:
o Jupyter Notebook (for experimentation and visualization)
o VS Code (for coding and development)
o Git (for version control)

Hardware Requirements

 Processor: Intel i7 or higher

 RAM: 8GB or more
 GPU: NVIDIA GTX 1060 or higher (for deep learning model training)
 Storage: At least 500GB of hard disk space (for dataset storage and model
weights)
 Camera/Video Streaming Device: For real-time video feed
 Operating System: Windows/Linux

Users

 Security Personnel: Monitoring surveillance footage.

 Law Enforcement: Detecting violence in real-time incidents.
 Researchers/Analysts: Studying patterns of violence.

Questionnaire for User Feedback

 Q1: How reliable do you think the system will be in detecting violence?
 Q2: What kind of video sources would you use the system on (CCTV,
mobile videos, etc.)?
 Q3: How important is real-time detection in your use case?
 Q4: What accuracy levels would you consider acceptable for this system?
 System Designs & Specification

High-Level Design (HLD)

The high-level design of the system includes the following components:

 Input Video: Real-time video streams or stored video files are fed into the
system.
 Preprocessing: Video frames are processed to detect objects and actions.
 Deep Learning Model: The core of the system is a deep neural network
trained to identify violence.
 Post-processing: Output involves labelling frames with violence or non-
violence and providing alerts.
 User Interface: Provides feedback and control over the system.

Project Model

The model could be based on Convolutional Neural Networks (CNN) or a

combination of CNNs with Recurrent Neural Networks (RNNs) for sequential data
analysis. A model like 2D CNN or 3D CNN would be suitable for spatial and
temporal feature extraction in video frames.
Structure Chart

A structure chart depicts the organization of the system:

pgsql
Copy
+----------------------+
| Video Input |
+----------------------+
|
+----------------------+
| Preprocessing Layer |
+----------------------+
|
+----------------------+
| Deep Learning Model |
| (CNN/RNN) |
+----------------------+
|
+----------------------+
| Post-Processing Layer|
+----------------------+
|
+----------------------+
| User Interface |
+----------------------+

Data Flow Diagram (DFD)

Level 1 DFD showing the flow of data within the system:

less
Copy
[Video Stream] ---> [Preprocessing] ---> [Deep Learning Model] ---> [Violence Detection]
| |
+------------------------------> [Alert/Notification]
Entity-Relationship (E-R) Diagram

The E-R diagram for the violence detection system:

csharp
Copy

[User] --<has>-- [Video Stream] --<processed by>-- [Model]

| |
[Violence Label] [Alert Notification]

UML Use Case Diagram

Use Case Diagram showing the interactions between the system and users:

pgsql
Copy
+------------------+
| User |
+------------------+
|
v
+-------------------------+
| View Video Feed |
+-------------------------+
|
v
+-------------------------+
| Receive Alerts |
+-------------------------+
|
v
+-------------------------+
| Analyse Violent Actions |
+-------------------------+
Low-Level Design (LLD)

Screen-Shot Diagram

Screenshot 1: Example of the user interface with video feed and alerts.

Screenshot 2: System output indicating detected violence, with a bounding box or

alert overlay
Coding
Testing
 CONCLUSION

The violence detection system using deep learning has significant applications in
improving safety in public places, monitoring law enforcement situations, and
assisting security personnel in real-time decision-making. By utilizing CNNs, we
can process video frames effectively to identify violent actions.

Limitations

 Accuracy: Detection may fail under certain lighting conditions or with

complex backgrounds.
 Real-Time Processing: Requires significant computational resources for
real-time video analysis.
 Data Availability: The system’s effectiveness depends on the quality and
quantity of the training data.
 References
[1] A. M. Ali, S. S. S. R. Depuru, and R. V. D. D. Kumar, Proceedings of the 2021
IEEE International Conference on Data Science and Advanced Analytics (DSAA),
2021.
[2] A. Gupta, D. S. Chowdhury, and R. Mahapatra, Journal of Computer Vision
and Image Processing, vol. 6, no. 2, pp. 112-125, 2020.
[3] R. N. Jena, S. P. Maji, and D. L. Dey, Proceedings of the IEEE International
Conference on Image Processing (ICIP), 2020.
[4] A. Khan, T. N. Nguyen, and R. K. B. R. Reddy, Journal of Visual
Communication and Image Representation, vol. 60, pp. 121-132, 2019.
[5] P. H. Nguyen, H. M. Le, and Q. V. H. Le, Proceedings of the 2018
International Conference on Image Processing and Vision Engineering (IPVE),
2018.