UCLALibrary/mgap

Experimental data pipeline for testing feature recognition capabilities of various ML models on images from UCLA Library digital collections.

Machine Generated Annotations Pipeline

Installation

1. Install Docker and Docker Compose.

2. Download and extract this repository.

3. Obtain access to Amazon Rekognition and Clarifai.

4. Stand up and configure instances of the following:

   • Blacklight (at least one instance)
   • IIIF image server
   • Elucidate WebAnnotation server
   • RabbitMQ

5. Fill in the blanks in .env.

6. Fill in the blanks and replace the dummy configuration data in mgap.util.get_config.

7. Bring up the containers on your Docker host:

   docker-compose up -d

Testing

At the project root directory, run:

$ pytest

Usage

Terminal 1

1. Bring up the containers on your Docker host, if you haven't already:

   docker-compose up -d

Terminal 2

1. Install Python 3.5, or 3.6, and a Python virtual environment manager.

2. In the repository directory, create and activate a virtual environment:

   # GNU/Linux
   python -m venv venv_mgap
   . venv_mgap/bin/activate

3. Install dependencies:

   $ pip install -r requirements.txt

4. Move the example send script to the repository directory, so it can find the mgap package as an absolute import:

   $ mv examples/pika/send_messages.py .

5. Pipe some JSON to the example send script:

   $ echo '{ "iiif_image_info_url": "https://stacks.stanford.edu/image/iiif/gp903kf9548%2FSC1041_SAIL_Office_1979", "iiif_manifest_url": "https://purl.stanford.edu/gp903kf9548/iiif/manifest", "item_ark": "ark:/00000/aaa.bbb" }' | ./send_messages.py