CVR COLLEGE OF ENGINEERING

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

A Mini Project seminar

on
GUNS DETECTION SYSTEM USING DEEP LEARNING
by
P.Manmadha (20B81A0325)
B.Manoj Kumar(20B81A0326)
Under guidance of
S.Kasturi(Assistant professor )

1
Contents
1. Abstract
2. Motivation
3. Literature review
4. Limitations of the existing models
5. Proposed problem statement
6. Proposed approach of the work
7. User/Stakeholder requirement analysis
8. Software requirement/Technology stack
9. Hardware requirements
10.Data sets requirement
11. Innovation in the project
12. Plan of action to complete the project

2
Abstract
In the contemporary security landscape, the need for advanced technologies to enhance
threat detection is paramount. This research addresses the crucial task of weapon detection
through the development of a robust system employing deep learning algorithms. The
proposed system utilizes a dataset comprising images and videos with both weapon and
non-weapon instances. The methodology involves pre-processing the dataset to ensure
optimal compatibility with machine learning models, incorporating techniques such as
image augmentation and normalization. Feature extraction is conducted to identify relevant
patterns and characteristics in weapon-related imagery. Various deep learning models,
including convolutional neural networks (CNNs) are implemented and rigorously evaluated
to determine the most effective approach. Performance metrics such as precision, recall,
and F1 score are employed to assess the system's accuracy and efficiency in detecting
weapons while minimizing false positives.

3
Motivation
 The rising concern for public safety and the increasing
frequency of security threats in crowded areas motivate the
development of an advanced weapon detection system. By
leveraging deep learning, the project aims to provide a
proactive and efficient solution to enhance security
measures.

4
Literature review
 Grega et al. presented an algorithm which automatically detects
knives and firearms in CCTV image and alerts the security guard
or operator . Majorly, focusing on limiting false alarms and
providing a real-time application where specificity of the
algorithm is 94.93%.
 Ji et al. developed a system for security footage which
automatically identifies the human behavior using
convolutional neural nets (CNNs) by forming deep learning
model which operates directly on the raw inputs .
5
Limitations of the existing Applications/ Models

 Current weapon detection models may exhibit limitations

such as high false positives, slow processing times, or
challenges in adapting to diverse environments. Addressing
these limitations is crucial for the successful implementation
of an effective system.

6
Proposed problem statement

 The project aims to improve the accuracy and efficiency of

weapon detection systems, addressing specific limitations
identified in current models. The focus is on developing a robust
solution that can operate in real-time scenarios with minimal
false positives.

7
Proposed approach of the work

 The project will employ a Deep learning algorithms and

computer vision techniques for accurate weapon detection.
Data preprocessing, feature extraction, and model training will
be optimized to achieve high performance

8
User/Stakeholder requirement analysis

 Understanding the needs and expectations of end-users and

stakeholders, such as security personnel and administrators, is
crucial. The system should be user-friendly, provide real-time
alerts, and integrate seamlessly into existing security
infrastructure.

9
Software requirement/Technology stack

 PyCharm
 Python
 NumPy
 OpenCV
 YOLO v3
 Pygame

10
Hardware requirements

 Storage-512GB or more

 RAM-4GB or above

 Processor-Processor with at least 2.4GHz

11
Architecture of the Model
Input video Detected guns
Detected frame

Frame conversion
Training data Get location

Image preprocessing
Classification
Alert

Object detection

Database
Object identification

Object detection Analysis Action

12
Data sets requirement
 Different datasets were made for classification and detection. For real-time
purposes, we made our dataset by taking weapon photos from the camera,
data was extracted manually from robbery CCTV videos etc.

 Weapon Detection Dataset From Kaggle.

13
Modules in the Code

 NumPy
 OpenCV
 Pygame

14
Main functions and its implementation in the Code
for video capture Showing information on the screen
cap = cv2.VideoCapture(value()) class_ids = []
confidences = []
while True: boxes = []
x, vid = cap.read() for out in outs:
if not x: for detection in out:
print("Error: Failed to read a frame from the video scores = detection[5:]
source.") class_id = np.argmax(scores)
break confidence = scores[class_id]
height, width, channels = vid.shape if confidence > 0.5:

Detecting objects #Displaying the Result

blob = cv2.dnn.blobFromImage(vid, 0.00392, (416, cv2.imshow("Video", vid)
416), (0, 0, 0), True, crop=False) key = cv2.waitKey(1)
if key == 27:
net.setInput(blob) break

outs = net.forward(output_layer_names)

15
Innovation in the project

 The project introduces innovation through the integration of state-of-the-

art Deep learning learning techniques, ensuring a high level of accuracy in
weapon detection. The application of these methodologies to real-world
security scenarios represents a novel and impactful contribution.

16
Plan of action to complete the project

Week 1-2: Project Kick-off and Research

Week 3-4: Data Collection and Preprocessing
Week 6-7: Model Training and Evaluation
Week 7-8:Testing and Modifications
Week 9: Finalization and Presentation Preparation

17
References
 Narejo S, Pandey B, Esenarro Vargas D, Rodriguez C, Anjum MR. Weapon detection using
YOLO V3 for smart surveillance system. Mathematical Problems in Engineering. 2021 May
11;2021:1-9.

 M. Grega, A. Matiolanski, P. Guzik, and M. Leszczuk, “Au- ´ tomated detection of firearms and knives in a
CCTV image,” Sensors, vol. 16, no. 1, p. 47, 2016.

 Yadav P, Gupta N, Sharma PK. A comprehensive study towards high-level approaches for
weapon detection using classical machine learning and deep learning methods. Expert Systems
with Applications. 2023 Feb 1;212:118698.

 Jain H, Vikram A, Kashyap A, Jain A. Weapon detection using artificial intelligence and deep
learning for security applications. In2020 International Conference on Electronics and Sustainable
Communication Systems (ICESC) 2020 Jul 2 (pp. 193-198). IEEE.

18