What is SIS used for?

SIS is a simple Python image search engine using Keras and Flask: it extracts VGG16 features offline and finds similar images with a linear scan. It is a good fit for teaching image retrieval basics and prototyping visual search for small collections.

Yes. SIS is open source under the MIT licence. Its source code is on GitHub at matsui528/sis and is written mainly in Python.

Yes. SIS is open source, so the software itself is free to use.

Yes. SIS can be self-hosted on your own server or infrastructure; there is no official hosted version.

What are some alternatives to SIS?

Similar open-source tools in the Search category include Trace.moe, SearXNG and SearX. SaaS products in the same category include Bing, Yandex Search and Baidu.

Is SIS actively maintained?

The most recent commit to SIS was on 14 November 2021. The project has 778 stars on GitHub.

SIS: open-source search

About SIS

Simple Image Search Engine, or SIS, is a small Python project for searching a collection of images by example. It uses Keras and Flask, and you can start it by running just two Python scripts, which makes it a useful teaching and prototyping tool.

The offline script extracts a deep feature from every database image, a 4096-dimensional fully connected layer activation from a VGG16 model with ImageNet weights. The server script then starts a Flask web interface, where you upload a query image and the server finds similar images through a simple linear scan. No GPU is required, and it was tested on Ubuntu 18.04 and WSL2.

The README explains how to launch it on an AWS EC2 instance, with port 5000 opened, and suggests uWSGI plus nginx for a more secure deployment or AWS App Runner for a serverless setup. It is MIT licensed and was updated for a CVPR 2020 course, with slides and video linked.

Key features

Image-by-example search
VGG16 deep feature extraction with Keras
Flask web interface for queries
Linear scan similarity search
Runs without a GPU
AWS EC2 deployment instructions

Good fit for

→Teaching image retrieval basics
→Prototyping visual search for small collections

Built with: Python; Flask
Tags: image-search; image-retrieval; keras; flask; python; vgg16; search-engine; computer-vision

SIS: questions and answers

What is SIS used for?: SIS is a simple Python image search engine using Keras and Flask: it extracts VGG16 features offline and finds similar images with a linear scan. It is a good fit for teaching image retrieval basics and prototyping visual search for small collections.
Is SIS open source?: Yes. SIS is open source under the MIT licence. Its source code is on GitHub at matsui528/sis and is written mainly in Python.
Is SIS free?: Yes. SIS is open source, so the software itself is free to use.
Can I self-host SIS?: Yes. SIS can be self-hosted on your own server or infrastructure; there is no official hosted version.
What are some alternatives to SIS?: Similar open-source tools in the Search category include Trace.moe, SearXNG and SearX. SaaS products in the same category include Bing, Yandex Search and Baidu.
Is SIS actively maintained?: The most recent commit to SIS was on 14 November 2021. The project has 778 stars on GitHub.