System to Detect Fake Logos Online
Final Year Project Report
Submitted in partial fulfillment of the requirements for the degree of Bachelor of Computer
Applications
Submitted by:
[Your Name]
Roll No: [Your Roll Number]
College Name: [Your College Name]
Under the guidance of:
[Guide Name]
Date: April 20, 2025
Abstract
This project presents a system to detect fake logos online using image processing and machine
learning techniques. With the rise in counterfeit branding, this project provides a reliable method to
identify logo authenticity.
Introduction
Brand counterfeiting has become a major issue across various industries. The internet has enabled
easy circulation of duplicate logos. An automated detection system can help brands, designers, and
platforms stay alert.
Problem Statement
There is a need for an automated system to detect counterfeit logos from images found online.
Objectives
- Build a logo detection and classification system.
- Compare logos against a dataset of original and fake logos.
- Provide accuracy metrics.
- Output whether a logo is real or fake.
Existing System
Manual logo verification is time-consuming and inconsistent. Some systems use watermarking or
digital signatures, but they're not effective for general image searches.
Proposed System
The proposed system uses a CNN model trained on a dataset of logos to automatically detect and
classify whether a logo is fake or real.
Technologies Used
- Python (for training model and backend logic)
- TensorFlow/Keras (for CNN)
- HTML, CSS (for front-end interface)
- Java/C++ (for optional integration modules, image preprocessing)
System Design
The system follows a flow: Image input -> Preprocessing -> Feature Extraction -> Classification ->
Output.
Implementation
1. Data Collection:
- Downloaded and labeled images of real and fake logos.
2. Image Preprocessing (in Java/C++):
- Converted images to grayscale, resized them.
3. Model Training (Python):
```python
from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense
model = Sequential([
Conv2D(32, (3,3), activation='relu', input_shape=(64,64,3)),
MaxPooling2D(pool_size=(2,2)),
Flatten(),
Dense(128, activation='relu'),
Dense(1, activation='sigmoid')
])
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
```
4. Front-End (HTML/CSS):
```html
<form action="/upload" method="POST" enctype="multipart/form-data">
<input type="file" name="logo">
<button type="submit">Check Logo</button>
</form>
```
Output Screenshots
Sample output: 'The uploaded logo is likely FAKE with 89% confidence.'
Advantages
- Fast detection
- User-friendly interface
- Scalable model
Limitations
- Needs a large, diverse dataset
- Struggles with highly distorted images
Future Scope
- Add support for video/logo detection in motion
- Improve model using transformers
- Build mobile app integration
Conclusion
This project successfully builds a foundational system for detecting fake logos using AI. It can be
further expanded into commercial tools or integrated with brand security software.
References
- https://keras.io/
- https://opencv.org/
- Research papers on logo detection using CNN
- GitHub repositories for public datasets
