https://github.com/1adore1/logo-similarity
A web app for logo similarity detection using deep learning. Upload a test logo and multiple reference logos to compare them using feature embeddings extracted with a ResNet-50 model.
https://github.com/1adore1/logo-similarity
python pytorch streamlit
Last synced: 2 months ago
JSON representation
A web app for logo similarity detection using deep learning. Upload a test logo and multiple reference logos to compare them using feature embeddings extracted with a ResNet-50 model.
- Host: GitHub
- URL: https://github.com/1adore1/logo-similarity
- Owner: 1adore1
- Created: 2025-02-02T14:14:24.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-03-26T06:28:08.000Z (over 1 year ago)
- Last Synced: 2025-03-26T07:29:27.273Z (over 1 year ago)
- Topics: python, pytorch, streamlit
- Language: Jupyter Notebook
- Homepage: https://logo-similarity.streamlit.app/
- Size: 90 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Logo Similarity
This project demonstrates a system that, given a cropped region of an image (e.g., a logo extracted from a video with embedded advertising), answers the question:
**"Is this the logo of the target organization?"**
The system compares the provided test logo against several reference logo samples of the organization. It is designed to handle logos with varying scales, brightness, contrast, rotations, and minor design variations. The underlying idea is to use a deep learning model to generate embeddings for each logo and then compute the similarity (via cosine similarity) to decide if the test logo matches one of the provided samples.
Try it: [logo-similarity.streamlit.app](https://logo-similarity.streamlit.app/)
## How It Works
1. **Upload Reference Logos:** Users can upload multiple images representing the target organization's logos.
2. **Upload Test Logo:** Users upload a cropped image containing the test logo.
3. **Comparison:** The system processes each image through a deep learning model, computes embeddings, and calculates cosine similarity between the test logo and each reference.
4. **Result Display:** The system displays the similarity score and a match/no-match result based on a predefined threshold.
## Getting Started
### Requirements
- Python 3.x
- [Streamlit](https://streamlit.io/)
- PyTorch
- PIL
### Installation
1. Clone the repository:
```
git clone https://github.com/1adore1/logo-similarity.git
cd logo-similarity
```
2. Install the required packages:
```
pip install -r requirements.txt
```
### Running the Application
Launch the app using Streamlit:
```
streamlit run src/main.py
```
Then, open the provided URL in your browser to interact with the system.
## Tech Stack
- **Frontend**: Streamlit
- **Backend**: Python
- **ML Model**: `ResNet50`