Image Retrieval Project

Overview

This project is focused on developing an image retrieval system that identifies and retrieves similar images from a dataset. The system employs various image processing techniques and similarity measures to achieve efficient and accurate retrieval.

Table of contents

• Installation
• Usage
• Project Structure
• Methodology
• Results

Installation

Prerequisites

Ensure you have the following packages installed:

• numpy
• PIL
• matplotlib

Set up

1. Clone the repository

git clone https://github.com/K1n031/Image_Retrieval_Project
cd Image_Retrieval_Project

2. Download dataset Use the provided script to download and unzip the dataset.

!gdown 1msLVo0g0LFmL9-qZ73vq9YEVZwbzOePF
!unzip data.zip

Usage

1. Repare dataset The dataset should be organized into a folder structure where each class of images has its own folder.

data/
├── test/
│   ├── class_1/
│   ├── class_2/
│   └── ...
├── train/
│   ├── class_1/
│   ├── class_2/
│   └── ...

2. Run the Project Open the Jupyter notebook and execute the cells step by step to process the images, extract features, and perform image retrieval.

jupyter notebook Image_Retrieval_Project.ipynb

Project Structure

• 'Image_Retrieval_Project.ipynb': The main notebook containing the code and explanations.
• 'data/': Directory where the image dataset is stored.
• 'requirements.txt': List of dependencies needed for the project.

Methodology

1. Image Processing

• Image Resizing: All images are resized to a uniform size of 448x448 pixels.
• Feature Extraction: Custom functions are used to convert images into a numerical format suitable for comparison.

2. Similarity Measures

• L1 Measure (Absolute Difference): The primary metric used to compare images based on pixel differences.

3. Visualization

• Result Plotting: The system displays the query image alongside the top retrieved images from the dataset for visual inspection.

Results

The system effectively retrieves similar images from the dataset, showcasing its ability to identify images with visual similarities based on the chosen metrics.