Fake Logo Detection DT Report
Fake Logo Detection DT Report
Submitted By
RAJALAKSHMI.R 713522CS801
Submitted to the
of
MASTER OF ENGINEERING
FEBRAURY 2023
i
BONAFIDE CERTIFICATE
Examiner 1 Examiner 2
THE IMPORTANCE OF DESIGN THINKING-AN OUTLINE
1 Problem Statement 1
2 Empathy 2
3 Define 4
4 Ideate 5
5 Prototype 9
6 Testing 15
LIST OF FIGURES
LIST OF ABBREVIATIONS
PROBLEM STATEMENT
1
CHAPTER-2
EMPATHY
To help the people by making them buy only the original products and
also help the brands to maintain their standard. To help the brands to know where
the fake products are being sold and helps to find them out and take sufficient
action on that case. Finally, it helps every people to buy quality and original
products rather than the fake one.
2
ORGANIZATION OF THE PROJECT REPORT
3
CHAPTER-3
DEFINE
Defining a project, along with its goals, objectives, and key deliverables
clarifies and aligns project expectations for all involved parties. Agreement on
important project details signals the transition to begin the project work.
The existing system provides the way to find only the brand logo so the
system is not so useful to prevent people from buying forged projects. The
existing system can find only the logo but can’t find whether its real or fake. Since
the differences is more minute people can’t identify it easily. In The Existing
System The just the logo name is displayed with the grid like structure around the
logo and not much thing to find the fake one. The old systems that didn’t used
the yolo very much struggled to find the logo in the real time environment. Since
the old systems scanning speed is less and it doesn’t have the ability to find the
small logo images.
The proposed fake Logo Detection System aims to help consumers
distinguish fake from the original product. Using this system, a consumer can
verify whether a product real or forged. This System can also be helpful for
brands struggling to fight against forged products. This system allows the users
to complain the original brand about the fraudulent activities, so that the brand
owners can take sufficient action to reduce the fraudulent activities and prevent
the people from buying the fake products and also they can prevent themselves
from damaging their credibility.
4
CHAPTER-4
IDEATE
In this system the input is the logo of the product detected in real-time and
the output is whether the logo is fake or real. The system first compares the input
with the trained dataset from the admin side. For comparing the images in the
dataset the Tensorflow module is used which is one of the main model of the deep
learning library. For real time input Open Cv is used. Open Cv is the python
library helps to access the camera. For the real time detection and analyzing the
originality of logo the Yolo algorithm is been used.
In future it can be developed in a way users can complain the original brand
about the fraudulent activities, so that the brand owners can take sufficient action
to reduce the fraudulent activities and prevent the people from buying the fake
products and also they can prevent themselves from damaging their credibility.
Overall system architecture for the proposed system is shown in Figure 4.1
as below.
5
Fig 4.1 Logo Detection Architecture Diagram
6
USE CASE DIAGRAM
The use case for the proposed system. System modeling has been done
through a use case that represents a user's interaction with the system that shows
the relationship between the user and the different use cases in which the user is
involved.
The overall system architecture for the proposed system is shown in Figure
4.2 as below
7
In this chapter, the system is divided into various modules for better
understanding and the working of several aspects of the application. Also, the
UML diagrams are drawn up for the proposed system to highlight the interactions
between the modules and to analyze the efficiency of their capabilities. As a
result, the proposed Augmented Reality app is secure and can streamline effective
and efficient work for the Customers.
8
CHAPTER-5
PITCHING PROTOTYPE
For the implementation of the proposed system, the The Python has been
chosen for several reasons. First, it provides all the necessary tools to create
interactive environments in a relatively simple way. Second, it has a complete
and well documented. And finally, it can run on different platforms such as PCs,
consoles, mobile devices or web.
Hardware Requirements
Processor: Intel i5/i7/i9(8th Gen or above)
Download speed of 10 Mbps or greater
RAM: 8 GB (or above)
Hard disk: 20 GB (or above)
Software Tools Required
Python 3.7 or above
Tensorflow
OpenCV
TECHNOLOGIES USED
OPENCV
9
OpenCV is the open-source library for the computer vision, machine
learning, and image processing and now it plays a major role in real-time
operation which is very important in today’s systems. By using it, one can process
images and videos to identify objects, logos, faces, or even handwriting of a
human. When it integrated with various libraries, such as NumPy, python is
capable of processing the OpenCV array structure for analysis. To Identify image
pattern and its various features we use vector space and perform mathematical
operations on these features. When OpenCV was designed the main focus was
real-time applications for computational efficiency.
OpenCV Functionality
1. Image/video I/O, processing, display
2. Object/feature detection
3. Geometry-based monocular or stereo computer vision
4. Computational photography
5. Machine learning & clustering
6. CUDA acceleration
Applications of OpenCV:
1. face recognition
2. Automated inspection and surveillance
3. number of people – count (foot traffic in a mall, etc)
4. Vehicle counting on highways along with their speeds
5. Interactive art installations
6. Anamoly (defect) detection in the manufacturing process (the odd
defective products)
7. Street view image stitching
8. Video/image search and retrieval
9. Robot and driver-less car navigation and control
10. object recognition
11. Medical image analysis
10
12. Movies – 3D structure from motion
13. TV Channels advertisement recognition
TENSORFLOW
11
execution and the accuracy in placing bounding boxes.
Tensorflow bundles together Machine Learning and Deep Learning
models and algorithms. It uses Python as a convenient front-end and runs it
efficiently in optimized C++.Tensorflow is at present the most popular
software library. There are several real-world applications of deep learning
that makes TensorFlow popular. Being an Open-Source library for deep
learning and machine learning, TensorFlow finds a role to play in text-based
applications, image recognition, voice search, and many more. DeepFace
application Facebook’s image recognition system, uses TensorFlow for image
recognition. It is used by Apple’s Siri for voice recognition. Every Google
app that you use has made good use of TensorFlow to make your experience
better.
YOLO
The YOLO framework (You Only Look Once) on the other hand, deals
with object detection in a different way. It takes the entire image in a single
instance and predicts the bounding box coordinates and class probabilities for
these boxes. The biggest advantage of using YOLO is its superb speed – it’s
incredibly fast and can process 45 frames per second. YOLO also understands
generalized object representation.
This is one of the best algorithms for object detection and has shown a
comparatively similar performance to the R-CNN algorithms.
Compared to the approach taken by object detection algorithms before
YOLO, which repurpose classifiers to perform detection, YOLO proposes the use
of an end-to-end neural network that makes predictions of bounding boxes and class
probabilities all at once.
Following a fundamentally different approach to object detection, YOLO
12
achieves state-of-the-art results beating other real-time object detection algorithms
by a large margin.
In addition to increased accuracy in predictions and a better Intersection over
Union in bounding boxes (compared to real-time object detectors), YOLO has the
inherent advantage of speed.
While algorithms like Faster RCNN work by detecting possible regions of
interest using the Region Proposal Network and then perform recognition on those
regions separately, YOLO performs all of its predictions with the help of a single
fully connected layer.
Methods that use Region Proposal Networks thus end up performing
multiple iterations for the same image, while YOLO gets away with a single
iteration.
The YOLO algorithm works by dividing the image into N grids, each having
an equal dimensional region of SxS. Each of these N grids is responsible for the
detection and localization of the object it contains.
13
SAMPLE LOGO DATASETS NAME
Bic
Samsung
Pepsi
Lays
Mars
MnM
Mtn dew
Oreo
Heinz
Marvel
PlayStation
Chevrolet
Burger King
Hp
Fila
Microsoft
Chrome
NASA
Reebok
Oral b
Cowbell
Peak milk
Twitter
Google
Adidas
Android
Nutella
Puma
Pringles
Sprite
Tesla
Netflix
Zara
Huawei
YouTube
Lipton
Nestle
14
CHAPTER-6
TESTING
TRAINING MODULE:
One of the best ways to improve deep learning performance is to improve
the data. Custom Dataset created with the help of sample images which then
augmented using data augmentation techniques to add more data to the dataset.
In the proposed system the logo dataset consists of two classes Positives and
Negatives.
The positive dataset consists of the whole set of images of the original
brands. Only original brands logo are present in the positives.There are multiple
images being collected for each and every brands for the sake of getting the
accurrate results.The intensities of images taken are grayscale and normal rgb
coloured images.
The Negative dataset consists of the whole set of images of the original
brands. Only Fake brand logos are present in the Negatives .There are multiple
images being collected for each and every brands for the sake of getting the
accurrate results.The intensities of images taken are grayscale and normal rgb
15
coloured images.
The data training can be carried out on the various training techniques
and the training time .the epoch time and the way of grouping the batches is very
much important while training the dataset.
An epoch is a term indicates the number of passes of the entire training
dataset the machine learning algorithm has completed.The data can be grouped
into multiple batches like 8,16, 32,64 etc.The no of batches is very much related
to the epoch time.If the number of batches are high the data training will be fast
and it consumes less epoch time. These were the the processes that happens while
training the dataset.
16
Fig 6.1b Training Model
17
The Detection Module
After the completion of the training module the detection takes place. In
this module the main aim is to detect the region of the logo.Once the training is
completed the model file is generated which is primarily used to compare the logo
detected in realtime with the logo in the dataset.
The primary component required for the detection is the camera . when the
camera is turned on the logo is shown infront of the camera and it starts detecting
it and it analyze the results about the logo.
These were the processes that takes place in the detection module of the fake logo
identification system.
18
THE OUTPUT MODULE
This is the final phase in which the output is shown to the user after
analyzing all the factors. The output is shown in a bounding box labelled as
“Original Brand logo” And “Fake Brand Logo”. So in this phase the output user
got will be useful for him to buy the original product and prevent themselves from
buying it.The system differentiate the original and fake ones by comparing with
the elements in the dataset.
19
FIG 6.3b Detecting Original Adidas Logo
20
CONCLUSION
21