This project is based on Deep Learning (ResNET50) and Machine Learning. It is similar to the google and amazon image search option, where you can upload any image and get the recommendation similar to the given image. In this project we use Myntra and Amazon image dataset from kaggle which include watches, shoes, colths. As an update we can train this model with other datasets also like furniture, cars, etc.

ResNET50 Convolutional neural network (CNN)
K-Nearest Neighbors Algorithm (KNN) Streamlit

📚 Documentation

Document	Description
Technical Specs	Architecture, technology stack, and API reference
Getting Started	Installation, setup, and running the application
Technical Debt	Suggested improvements and refactoring ideas

🏗️ Architecture Summary

The system uses a two-stage approach:

Feature Extraction: ResNet50 (pre-trained on ImageNet) extracts 2048-dimensional feature vectors from images
Similarity Search: K-Nearest Neighbors finds the most similar images using Euclidean distance

Interfaces:

Streamlit UI: Interactive web app for uploading and searching images
Flask API: REST endpoints for programmatic access

See docs/SPECS.md for detailed technical specifications.

🚀 Quick Start

Prerequisites

Python 3.8 - 3.11
8GB RAM minimum
~20GB disk space

Installation

git clone https://github.com/deepamkalekar/Reverse-Image-Search-ML-DL-Project.git
cd Reverse-Image-Search-ML-DL-Project
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt

Setup Dataset

Download Fashion Product Images Dataset (~16GB)
Create images/ folder and extract dataset images there
Create required directories: mkdir -p uploads embeddings filenames

Generate Embeddings

python run_generate_embeddings.py

Run the Application

Streamlit Web UI:

streamlit run run_streamlit_ui.py

Access at http://localhost:8501

Flask API:

python run_api_server.py

API available at http://localhost:5002

See docs/GETTING_STARTED.md for detailed instructions, troubleshooting, and configuration options.

Run Locally (Legacy)

Clone the project

  git clone https://github.com/deepamkalekar/Reverse-Image-Search-ML-DL-Project.git

Go to the project directory

  cd Reverse-Image-Search-ML-DL-Project

Download Dataset :- Product Image Data (16 GB)
Create folder images and move all the download images to this folder
Create folder uploads in the same directory for user, whenever user upload a photo it will be saved in this folder

Install dependencies

  pip install -r requirements.txt

Generate Embeddings

python run_generate_embeddings.py

Run the Streamlit Web App

streamlit run run_streamlit_ui.py

Demo

Click On Browse File and upload image