Image Search Engine

This project is an image search engine that allows users to upload an image and find similar images from a preloaded database. It uses the ResNet50 model for feature extraction and cosine similarity for finding similar images.

Features

• Image Upload: Users can upload an image in JPG format.
• Feature Extraction: Extracts features from the uploaded image using a pre-trained ResNet50 model.
• Similarity Search: Finds similar images from the database using cosine similarity.
• Adjustable Parameters: Users can adjust the similarity threshold and the number of similar images to display.
• Caching: Utilizes Streamlit's caching for improved performance.

Prerequisites

• Python 3.10
• pip (Python package installer)

Installation

1. Clone the repository:

   git clone https://github.com/nasirovsh/ecommerce-visual-search.git
   cd ecommerce-visual-search

2. Create a virtual environment:

   python3 -m venv .venv
   source .venv/bin/activate  # On Windows use `.venv\Scripts\activate`

3. Install the required packages:

   pip install -r requirements.txt

   For better performance, install the Watchdog module:

   For macOS users:

   xcode-select --install
   pip install watchdog

   For Windows users:

   pip install watchdog

Usage

1. Prepare the image database:

   • Place your database images in the db/ directory.
   • Ensure all images are in JPG format.

2. Run the Streamlit app:

   • Run in terminal:

     streamlit run main.py

   • Run/Debug in PyCharm:

     Go to Run/Debug Configurations and add a new configuration for Streamlit Server. Set module name to streamlit and parameters to run main.py.

     Click on the Run button to start the Streamlit server.

   • Open the browser and go to the URL displayed in the terminal (usually http://localhost:8501).

3. Use the app:

   • Upload an image: Use the file uploader to select an image in JPG format.
   • Adjust parameters: Use the sliders to set the similarity threshold and the number of similar images to display.
   • Find similar images: Click the "Find Similar Images" button to search for similar images in the database.

Project Structure

• main.py: The main script that runs the Streamlit app.
• requirements.txt: Lists the dependencies required for the project.
• db/: Directory containing the image database.
• README.md: This file, containing project documentation.
• LICENSE: The license file for the project.

Dependencies

• pillow == 10.3.0: Python Imaging Library for opening, manipulating, and saving images.
• tensorflow == 2.16.1: Open-source machine learning framework used for the ResNet50 model.
• streamlit == 1.37.1: Framework for building interactive web applications.
• scikit-learn == 1.5.1: Machine learning library used for cosine similarity calculation.
• certifi == 2024.7.4: Provides Mozilla's carefully curated collection of Root Certificates.

How It Works

1. The app loads a pre-trained ResNet50 model on startup.
2. When a user uploads an image, the app extracts features using the ResNet50 model.
3. These features are compared to the pre-extracted features of images in the database using cosine similarity.
4. The app displays the most similar images based on the user-defined threshold and number of results.

Troubleshooting

• If you encounter memory issues, try reducing the number of images in your database or upgrading your hardware.
• Ensure your uploaded images are in JPG format and are not corrupted.
• If the app is slow, it might be due to the initial loading of the database. Subsequent runs should be faster due to caching.

Contributing

Contributions to this project are welcome. Please fork the repository and submit a pull request with your changes.

License

This project is licensed under the MIT License. See the LICENSE file for more details.

Acknowledgements

• The ResNet50 model is provided by TensorFlow and was originally developed by Microsoft Research.
• Thanks to the Streamlit team for their excellent framework for building data applications.