/content-based-image-search

Simple project that implements a content-based image retrieval engine using PostgreSQL as storage backend and Python for the application logic

Primary LanguagePythonMIT LicenseMIT

Content-based image retrieval

This is a simple project that implements a content-based image retrieval engine using PostgreSQL as storage backend and Python for the application logic. It uses VGG16 to extract features from images.

Dataset

You can find the original (Oxford-IIIT Pet Dataset) dataset used in this project in the following link. This dataset consists of roughly 7000 dog and cat images annotated with their respective breeds.

Requirments

Python

You will need to create a Python virtual environment and install the packages found in base_requirements.txt file.

Create and activate Python virtual env using anaconda

conda create --name image_search_env python=3.8
conda activate image_search_env

Install base required python packages

pip install -r base_requirements.txt

File base_requirements.txt contains all packages you need to install for this project except tensorflow related ones. This is because this project was developed in a Apple M1 machine and requires a different process to install tensorflow.

For Apple M1 do the following:

conda install -c apple tensorflow-deps
pip install tensorflow-macos
pip install tensorflow-metal

In other systems the following command should be enough (haven't tried tho):

pip install tensorflow

PostgreSQL

In order to use this application you need to set up a connection with a PostgreSQL database to create and populate a table named pet_images. To set up connection with PostgreSQL you need to create a .env file with the following format.

DB_HOST=localhost
DB_PORT=5432
DB_NAME=image_search
DB_USER=postgres
DB_SECRET=postgres

Use the Application

Follow the steps described in the next sections to setup and start the application.

Setting up dataset of images

  1. Download dataset of images using this link (clicking the link will start the downloading of a GZ file named images.tar.gz containing images folder).
  2. Unzip the images.tar.gz file to extract images folder
  3. Place images folder under data directory of the project

Setting up PostgreSQL database

  1. Execute the following command to create pet_images table that will host links to pet image files, metadata and feature vectors: 
    python db.py --action=create
  2. Execute the following command to ingest a sample of roughly 250 cat and 250 dog image records in pet_images table of PostgreSQL database. Images are randomly selected using take_image_files_sample from image_sampling.py module 
    python db.py --action=ingest

[Optional] Execute the following command to delete any image records in pet_images table:

python db.py --action=delete

Start streamlit web app

In order to start the Python web application activate the Python virtual environment as described in section Requirements and execute the following command

streamlit run app.py --server.port=5555

app_demo_gif