Content-based Image Retrieval System

Build content-based image retrieval system using deep learning, applied some large scale similarity search technicals like Kdtree, LSH, Faiss.

Demo

<img src="https://github.com/lehuutrung1412/ImageRetrieval/blob/main/docs/image_retrieval.gif" alt="demo" width="100%"/>

Usage

Usage:  python app/app.py [--large] [--feature] [--pca]
        Run demo app.
Options:
  --large=['kdtree', 'lsh', 'faiss']    Large scale method
  --feature=[PATH]                      Features indexing file path
  --pca=[INT]                           Enable pca

Usage:  python setup/export_feature.py [-path]
        Export feature indexing file to path.
Options:
  -path=[PATH]    Path to save features file

Usage:  python setup/evaluation.py [--large] [--feature] [--pca] [--top]
        Run system evaluation.
Options:
  --large=['kdtree', 'lsh', 'faiss']    Large scale method
  --feature=[PATH]                      Features indexing file path
  --pca=[INT]                           Enable pca
  --top=[INT][REQUIRED]                 Number of ranked lists element

Run

Run with Google Colab or Jupyter Notebook (Recommend with Colab resources)

1. Upload ImageRetrieval.ipynb to your Colab
2. Run all cells and go to address like xyz.ngrok.io to use

Run with docker

Run from remote docker image

Pull lastest image from docker hub and run

docker pull lehuutrung1412/image-retrieval
docker run -d -p 5000:5000 lehuutrung1412/image-retrieval

Build and run from source

1. Clone sourcecode

git clone https://github.com/lehuutrung1412/ImageRetrieval.git

2. Build docker image and run.

docker build -t image-retrieval .
docker run -d -p 5000:5000 image-retrieval