Skip to content
SPECIFICTION
Get listed

Open Image Search

github.com

WebSelf-hosted
Visit github.com
Spec score 7.0 #19 of 21 in Reverse Image Search Software

What it is

Open Image Search is an open-source reverse image search prototype for graphical collections and museums working with digitized prints and paintings. Its hosted demo searches more than 20,000 prints from the Graphische Sammlung of ETH Zürich. Collections can clone the project and run retrieval locally with their own image data. The search pipeline extracts image features with VGG16 and uses a k-nearest neighbors model to locate similar images. Image metadata can be supplied in a CSV file using column names defined in the documentation. The web interface is designed for mobile devices and can send searches from a phone camera. A Django REST Framework backend reformats uploads in memory, sends them to the feature-extraction model, and retrieves metadata for top matches. The project uses technologies including Vue, Django, TensorFlow, scikit-learn, SQLite, and Nginx. It is available as self-hosted software or a hosted web application under the MIT license. The repository is in development, prototype-stage, and read-only after archival on September 1, 2026.

Spec sheet

Purpose
Open Image Search is a prototype reverse image search tool for finding digitized print metadata and supporting workflows at graphical collections and museums.Read 8 Oct 2026 · github.com
Intended users
The project is aimed at graphical collections and museums working with digitized prints and paintings.Read 7 Oct 2026 · github.com
Support
The README directs inquiries to the ETH Library Lab contact form.Read 8 Oct 2026 · github.com
License
The software is distributed under the MIT License, which permits use, copying, modification, distribution, and sale subject to its terms.Read 8 Oct 2026 · github.com
Headquarters
Zurich, SwitzerlandRead 28 Sep 2026 · github.com
Backend
A Django REST Framework API reformats uploaded images in memory, sends them to the feature extraction model, and uses the results to retrieve metadata for top matches.Read 8 Oct 2026 · github.com
Development status
The README describes the application as in development and in a prototype phase.Read 8 Oct 2026 · github.com
Hosted access
The README links to an online application where users can upload pictures to find visually similar prints from the Graphische Sammlung of ETH Zürich.Read 8 Oct 2026 · github.com
Hosted collection
The README says the hosted demo lets users search a database of more than 20,000 prints from the Graphische Sammlung of ETH Zürich.Read 7 Oct 2026 · github.com
Image handling
Uploaded images are reformatted in memory before a request is sent to the feature extraction model.Read 7 Oct 2026 · github.com
Local use
The project can be cloned and run locally with a user’s own image data, metadata, and installed software dependencies.Read 8 Oct 2026 · github.com
Metadata
Users can supply image metadata in a CSV file, with permitted column names described in the project documentation.Read 7 Oct 2026 · github.com
Mobile access
The web interface is designed to work on mobile devices and supports sending search requests directly from a phone camera.Read 7 Oct 2026 · github.com
Mobile interface
The Vue web interface is designed to work on mobile devices and can send search requests from a phone’s camera.Read 8 Oct 2026 · github.com
Project status
The repository describes the application as in development and says it is in a prototype phase.Read 7 Oct 2026 · github.com
Repository status
GitHub says the repository was archived by its owner on September 1, 2026, and is read-only.Read 8 Oct 2026 · github.com
Search method
The application extracts image features with VGG16 and uses a k-nearest neighbours model to find similar images.Read 8 Oct 2026 · github.com
Technology
The README lists TensorFlow, scikit-learn, Vue, Django, SQLite, and Nginx among the project’s libraries and frameworks.Read 8 Oct 2026 · github.com

Strong points and limits

Strong points

  • Hosted search across more than 20,000 prints
  • Self-hosting with your own image data
  • VGG16 and nearest-neighbor similarity search
  • Mobile web and phone-camera requests
  • MIT-licensed project

Limits

  • Prototype and in-development project
  • Repository is archived and read-only
  • Search focuses on similar images

Who it suits

Open Image Search is intended for graphical collections and museums handling digitized prints or paintings. It suits organizations able to use a hosted demo or run the project themselves.

Verdict

Open Image Search provides a specialized reverse-image workflow for cultural collections, with hosted and local options. Its prototype status and archived repository matter for organizations planning ongoing maintenance.

Where Specifiction places it

Alternatives to Open Image SearchAll alternatives →