Image Similarity Search

An AI-powered application that finds similar images using deep learning techniques.

Project Overview

This project implements an image similarity search using a pre-trained Vision Transformer (ViT) model. It allows users to upload an image and find visually similar images from a pre-defined dataset.

Examples of the search with UI

Key Features

• Utilizes a Vision Transformer (ViT) model for feature extraction
• Pre-extracted features for quick similarity comparisons
• Flask web application for easy interaction
• Supports various image formats (PNG, JPG, JPEG, GIF)

Technologies Used

• 
• 
• 
• 
• 

Setup and Installation

1. Clone the repository

   git clone https://github.com/beavishead/similar_images_search.git
   cd similar_images_search
   

2. Set up a virtual environment

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

3. Install dependencies

   pip install -r requirements.txt
   
   

4. Run the application

   python app.py
   

5. Access the application

   • Open a web browser and go to http://localhost:5000

Project Structure

• app.py: Main Flask application
• model.py: Contains the model loading and feature extraction logic
• feature_extractor.py: Script for pre-extracting features from the dataset
• model_weights.pth: Pre-trained model weights
• pre_extracted_features.npz: Pre-computed features for the dataset

Future Improvements

• Deploy the app on web. The issues I faced are the dataset of images are being compared to is too large for the free tiers og deployment platforms.

• Make more meaningful app. The current app was made primarily as a blueprint of how a ml-model maybe implemented in a web application.

• More meaningful tasks would be: -- scraping images from the web -- detecting the potentially explicit content -- adding the ability to label images with a certain textual category

• make the uploaded by users images disappear after some time