Disguised Image

Recognition Using
Deep Learning
Review

N. Manoj, 41130297
G. Kiran Kumar, 41130243
8th SEM, ECE
 Introduction
This project aims to develop an advanced DIR system leveraging deep learning
techniques to enhance the accuracy and reliability of recognition systems.

Traditional image recognition methods often struggle with high false positive and
false negative rates when confronted with disguised images, highlighting the need
for more robust solutions.
Deep Learning:
Neural Networks with many layers (deep neural networks) to model complex patterns and
representations in data, enabling tasks such as image and speech recognition, natural
language processing, and autonomous systems.
It involves training large-scale models on vast amounts of data, to achieve high accuracy and
performance in various predictive and analytical tasks.
Disguised Image Recognition (DIR) poses a significant challenge in the field of computer vision
and security due to the complexity of accurately identifying individuals whose appearances
have been altered through the use of makeup, masks, or other means of disguise.
Block Diagram
Actual Hardware Photo
The Raspberry Pi Family: A Legacy of
Innovation
• he Raspberry Pi Foundation has consistently pushed the boundaries of
affordable computing with its diverse range of boards.
• Each generation has brought improvements in processing power, memory,
connectivity, and features, catering to a wider range of users and projects.
• Generations:
Raspberry Pi Zero
Raspberry Pi 1 Model A/B
Raspberry Pi 2 Model B
Raspberry Pi 3 Model A+/B+
Raspberry Pi 4 Model B
 Evolution of the Raspberry Pi
• Raspberry Pi 1 Model A/B (2012): The pioneers! Introduced the world to the
concept of a low-cost, accessible computer for education and experimentation.
Featured a single-core processor and limited RAM.
• Raspberry Pi 2 Model B (2015): A significant leap forward with a quad-core
processor, dramatically improving performance. Opened up possibilities for more
demanding applications.
• Raspberry Pi 3 Model B/B+ (2016/2018): Brought built-in Wi-Fi and Bluetooth
connectivity, making networking much easier. Processor speed also increased.
The B+ model refined some features and added faster Ethernet.
• Raspberry Pi 4 Model B (2019): A major upgrade with a faster processor, more
RAM options, dual micro-HDMI ports, USB 3.0, and Gigabit Ethernet. Enabled
desktop-like performance in a compact form factor.
Raspberry Pi 4 Model B: Tiny, Mighty
Computer
• The Raspberry Pi 4 Model B is a
versatile, credit-card-sized computer
that empowers you to learn, create,
and explore.
• Perfect for beginners and experts
alike, it's ideal for programming,
media centers, home automation,
and countless other projects.
• Affordable and packed with
features, the Raspberry Pi 4 Model B
is your gateway to endless
possibilities.
 Capabilities
• Powered by a quad-core ARM Cortex-A72 processor, the Raspberry Pi
4 Model B delivers a significant performance boost.
• With up to 8GB of RAM, you can handle demanding tasks and
multitask with ease.
• Run various operating systems like Raspberry Pi OS, Ubuntu, and
Windows 10 IoT Core.
• Develop and execute a wide range of applications, from games and
multimedia software to productivity tools and custom solutions.
 Features
• Dual micro-HDMI ports enable you to
connect two monitors for enhanced
productivity and immersive experiences.
• Two USB 3.0 ports provide lightning-fast
data transfer speeds for external devices.
• Gigabit Ethernet ensures seamless wired
networking for smooth online activities.
• Built-in 802.11ac Wi-Fi and Bluetooth 5.0
offer fast and reliable wireless connectivity.
• The standard 40-pin GPIO header allows
you to connect sensors, actuators, and
other electronic components for your
projects.
 Specifications
Processor: Quad-core ARM Cortex-A72 (up to
1.8 GHz)
Memory: 1GB, 2GB, 4GB, or 8GB LPDDR4-
3200 SDRAM
Storage: microSD card slot for operating
system and data storage
Video & Audio: 2x micro-HDMI ports (up to
4K@60Hz), 3.5mm audio jack
Connectivity: Gigabit Ethernet, 802.11ac Wi-
Fi, Bluetooth 5.0USB: 2x USB 3.0 ports, 2x
USB 2.0 ports
Power: 5V DC via USB-C connector
Dimensions: 85mm x 56mm
Luminescence Boards for Raspberry
Pi
• Luminescence boards are add-on
peripherals designed to enhance
your Raspberry Pi projects with
vibrant, controllable lighting.
• These boards offer a convenient way
to integrate LEDs into your projects,
providing visual feedback, creating
stunning light displays, or adding a
touch of ambiance.
• Easy to connect and program,
Luminescence boards unlock a world
of creative possibilities.
 Features and Capabilities
• Wide range of LEDs: Luminescence boards typically feature multiple LEDs,
often addressable RGB LEDs, allowing for individual color control and
complex lighting patterns.
• Easy connection: Designed to connect directly to the Raspberry Pi's GPIO
(General Purpose Input/Output) pins, simplifying setup.
• Software support: Libraries and example code are usually available for
popular programming languages like Python, making it easy to control the
LEDs.
• Daisy-chaining: Some boards allow for daisy-chaining, enabling you to
connect multiple boards together to create even larger and more complex
lighting setups.
• Low power consumption: Efficient design ensures minimal power draw on
your Raspberry Pi.
 Possible application in Projects
• Ambient lighting: Create dynamic mood lighting for any room.
• Status indicators: Use LEDs to display the status of your Raspberry Pi or
other connected devices.
• Gaming peripherals: Build custom controllers or enhance your gaming
experience with interactive lighting effects.
• Wearable electronics: Integrate LEDs into clothing or accessories for
unique and eye-catching designs.
• Home automation: Use LEDs to provide visual alerts or feedback for
your smart home devices.
• Art installations: Create interactive light sculptures or artistic displays.
 Steps to use the luminescence
board
• Connect the board: Carefully connect the Luminescence board to the
designated GPIO pins on Raspberry Pi.
• Install software: Install the necessary software libraries and
dependencies for Python.
• Writing the code:
• Advanced programming: Explore more complex lighting patterns,
animations, and color sequences.
• Integration with sensors: Use sensor data to trigger lighting effects or
create interactive displays.
• Power up: Power on your Raspberry Pi
 USB Webcam
• USB webcams offer a versatile and
cost-effective way to add vision to
Raspberry Pi 4 Model B projects.
• We can build a security system, a
time-lapse camera, or exploring
computer vision, a USB webcam
can provide the eyes your Pi needs.
• With plug-and-play compatibility
and a wide range of options
available, getting started is easy.
 Applications
• Security and Surveillance: Create a security system with motion detection
and remote viewing capabilities.
• Time-lapse Photography: Capture stunning time-lapse videos of plants
growing, construction projects, or changing weather patterns.
• Computer Vision: Explore object detection, facial recognition, and other
computer vision applications.
• Video Streaming: Set up a live video stream for monitoring, remote presence,
or broadcasting.
• Robotics: Give robots the ability to "see" and interact with their environment.
• Creative Projects: Use webcams for stop-motion animation, interactive
installations, or unique art projects.
 Facial Feature Extraction
• Facial feature extraction is a crucial process in
computer vision that involves identifying and
measuring specific landmarks on a human face.
• These landmarks, also known as facial keypoints,
can include the corners of the eyes, the tip of the
nose, the corners of the mouth, and other
distinctive points.
• Extracting these features enables a wide range of
applications, from facial recognition and emotion
analysis to augmented reality and medical
diagnosis
 Techniques for Facial Feature
Extraction
• Landmark Detection: Algorithms like the
Viola-Jones algorithm or Haar cascades
can detect the presence of a face and
then locate specific landmarks within it.
• Active Shape Models (ASMs): ASMs use
statistical models of facial shapes to
iteratively refine the location of keypoints,
taking into account variations in pose,
expression, and lighting.
• Deep Learning: Convolutional Neural
Networks (CNNs) have revolutionized
facial feature extraction, achieving high
accuracy and robustness even in
challenging conditions. Pre-trained
models are readily available.
• Other Methods: Other techniques include
template matching, edge detection, and
geometric approaches.
Python – Image Collect code
def collect_images(label, img_path, number_of_images): cv2.imshow('frame', frame)
#rtsp_stream_address = key = cv2.waitKey(1)
"rtsp://admin:[email protected]:554/Streaming/Ch
if key == ord('q'):
annels/102/"
break
#cap = cv2.VideoCapture(rtsp_stream_address)
elif key == 13: # Enter key
cap = cv2.VideoCapture(0)
imagename = os.path.join(img_path,
print(f"Collecting images for {label}") f"{label}_{count}_{uuid.uuid1()}.jpg")
print("Press Enter to capture an image. Press 'q' to quit.") cv2.imwrite(imagename, frame)
count = 1 print(f"Image {count} captured.")
while count <= number_of_images: # Capture three images count += 1
ret, frame = cap.read() cap.release()
cv2.destroyAllWindows()
print("Data collection for " + label + " completed")
return label
Python – Image Training code
import sys if cv2.waitKey(1) & 0xFF ==
print(sys.version) ord('q'):
print(sys.executable) break
import os, cv2, time
cap.release()
print('Processing Detection...')
cap = cv2.VideoCapture(0) cv2.destroyAllWindows()
while True:
success, img = cap.read()
img = cv2.resize(img,(0,0),None,0.25,0.25)
cv2.imshow('PIcam', img)
 Plan Ahead
• Write codes to read disguised images
• Extract facial features from disguised images
• Compare with undisguised image
• Find the most matching data from available facial features
• Identify the disguised image
Contributed Papers

• Paper submitted on “Disguised Image Recognition – A review”