iART

Overview

The project iART is devoted to the development of an e-Research-tool for digitized, image-oriented research processes in the humanities and cultural sciences. It not only aims to improve the efficiency of retrieval in image databases but also offers various tools for analyzing image data, thereby enhancing scientific work and facilitating new theory formation. The motivation for the project stems from the fundamental importance of the comparative approach in art history, which targets the similarity of pictures and comes along with a rehabilitation of similarity thinking in contemporary philosophy of science. iART is supposed to transfer the approach of art history theorists and practitioners of Comparative Analysis to the digital age, and to extend it by virtue of modern information technology.

Installation

At a later point there will be a docker container provided here.

Development setup

Requirements

• docker
• docker-compose

Setup process

1. Clone the iART repository including submodules:

   git clone --recurse-submodules https://github.com/TIBHannover/iart.git
   cd iart

2. Run install.sh to download and extract models:

   bash install.sh  # CPU
   bash install.gpu.sh  # GPU

3. Build and start the container:

   sudo docker-compose up --build  # CPU
   sudo docker-compose -f docker-compose.gpu.yml up --build  # GPU

4. Apply database migrations and build frontend packages:

   sudo docker-compose exec backend python3 manage.py migrate auth
   sudo docker-compose exec backend python3 manage.py migrate
   sudo docker-compose exec frontend npm install --force
   sudo docker-compose exec frontend npm run build

5. For demonstration purposes, index ./data/examples/wikipedia_small.jsonl:

   sudo docker-compose exec indexer python -m iart_indexer --m client --task indexing --path /data/examples/wikipedia_small.jsonl --image_paths /data/media

   Wait until the docker-compose process finishes indexing (this may take a few minutes). After that, an index must be created for faster searching and all existing images must be imported into the new index.

   sudo docker-compose exec indexer python -m iart_indexer -m client --task faiss_train --port 50151  

   sudo docker-compose exec indexer python -m iart_indexer -m client --task faiss_indexing --port 50151 

6. Go to the frontend instance at http://localhost/.

Code reloading

Hot reloading is enabled for backend. To display frontend changes, run:

sudo docker-compose exec frontend npm run build

Alternatively, use serve to enable a hot reloaded instance on http://localhost:8080/:

sudo docker-compose exec frontend npm run serve

About the project

iART was funded by the DFG from 2019 to 2021. Our team consists of Matthias Springstein, Stefanie Schneider, Javad Rahnama, Ralph Ewerth, Hubertus Kohle, and Eyke Hüllermeier.

Contributing

Please report issues, feature requests, and questions to the GitHub issue tracker. We have a Contributor Code of Conduct. By participating in iART you agree to abide by its terms.