Metis image enhancer repository

Table of Contents

• Table of Contents
• About the Project
• Description
• Built With
• Getting Started
• Prerequisites
• Installation
• Usage
• Example
• Architecture
• Roadmap
• Contributing
• Acknowledgements

About the Project

Description

Library that calls python api from java. The library uses image enhancement (super resolution) on small thumbnails, and performs a simple sharpen filter.

Built With

• Python
• Java
• Test containers
• Junit
• Archunit

Getting Started

Prerequisites and Dependencies

• Docker Engine v24.0.4
• Docker Compose v2.19.1

Installation of library

Get the repository

git clone https://github.com/europeana/metis-image-enhancer.git

Build docker ISR API

go to the cloned folder of the library /metis-image-enhancer/metis-image-enhancer-python-rest/src/main, and run the following command to build the image enhancer application

docker compose build --no-cache

Usage

Enable Python API Endpoint for image enhancement

go to the cloned folder of the library /metis-image-enhancer/metis-image-enhancer-python-rest/src/main, and run the following commands to get the service up and running this command will run with 1 node

docker compose up 

the application uses a python api to enhance the image

1. Image enhancement is possible with the command:

curl --request POST http://localhost:8080/enhance/image --form image=@./metis-image-enhancer-client/src/test/resources/img/thumbnail.jpg \ 
                                                        --output ./metis-image-enhancer-client/src/test/resources/img/thumbnail_enhanced.jpg

Response:

HTTP code: 200 
Content-Type: image/jpeg
Response Body: Enhanced image to download

2. Client java library

Note: Java and Maven are prerequisites, the client adds sharpen filter extra to the python api result to improve the end result.

installation

mvn clean install

usage

add dependency in your project

<dependency>
    <groupId>eu.europeana.metis.image.enhancement</groupId>
    <artifactId>metis-image-enhancer-client</artifactId>
    <version>1</version>
</dependency>

properties

By default, the following properties are set in the library if you want to change the address or port, and the timeout are in seconds. the property worker.service.type can have the values api or script this indicates which approach will use to process the images.

api.url = http://localhost:8080
connect.timeout = 300
read.timeout = 300
worker.service.type = api

command line stand alone script

If you only want to use the image processing without the API, you can invoke this from the command line, using the script directly. Your python environment need to comply with the following dependencies (filetype, flask, waitress, uuid, pillow, tensorflow, and numpy)

python ./metis-image-enhancer-python-rest/src/main/mie.py --input image_input.jpg --output image_output.jpg

General limitations

Why this AI python model runs as a script? in this iteration the python model cannot be held in memory otherwise you will get an Out of Memory. The model in python of image processing doesn't free the resources every time. Steps of usage: load, process, get results, discard, repeat again.

See demo project

run docker image previous to execute the demo app and then execute the following command in the cloned folder of the /metis-image-enhancer/metis-image-enhancer-demo

mvn clean package

execute the app e.g.

java -jar ./target/metis-image-enhancer-demo-1-SNAPSHOT.jar

example of demo code can be found in the metis-image-enhancer-demo module.

Example

The following image shows the outcome:

Original

Enhanced

Design architecture

The motivation behind the design summarizes in the following concept: "Communication goes from outer layers to the inner layers."

• The client layer contains interactions details between the configuration and the domain.
• The configuration layer provides the properties needed to connect with the isr app.
• The domain layer uses the adapter pattern implementation to plug in any client adapter implementation to do the image enhancement. It contains the model and worker core business logic, so this remains untouched by external changes of the outer layers as much as possible.

Roadmap

• library initial release.
• restructure to metis style projects.
• use stream of bytes instead of file.
• ...

Contributing

Any contributions you make are appreciated.

1. Fork the Project
2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
3. Commit your Changes (git commit -m 'Add some AmazingFeature')
4. Push to the Branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

Acknoledgements

This project is powered by Image Super Resolution v2.2.0