https://github.com/eurohlam/grocery-expenses-pipeline

command-line pipeline that parses grocery receipts and converts them into key-value table
https://github.com/eurohlam/grocery-expenses-pipeline

java ocr opencv tesseract

Last synced: 4 months ago
JSON representation

command-line pipeline that parses grocery receipts and converts them into key-value table

• Host: GitHub
• URL: https://github.com/eurohlam/grocery-expenses-pipeline
• Owner: eurohlam
• License: gpl-3.0
• Created: 2025-02-05T00:59:03.000Z (4 months ago)
• Default Branch: main
• Last Pushed: 2025-02-18T00:07:41.000Z (4 months ago)
• Last Synced: 2025-02-18T01:22:26.520Z (4 months ago)
• Topics: java, ocr, opencv, tesseract
• Language: Java
• Homepage:
• Size: 7.46 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# grocery-expenses-pipeline

The idea of this project is to create a simple command-line pipeline that parses grocery receipts and convert them into 

key-value table that can be represented as a HashTable or a JSON

The project is implemented in Java 17 and uses: 

- [tesseract](https://github.com/tesseract-ocr/) as OCR (optical character recognition) engine

- [opencv4](https://opencv-java-tutorials.readthedocs.io/en/latest/) to process the receipt image in order to improve it for OCR.

Only [bilateralFilter](https://docs.opencv.org/4.x/dc/dd3/tutorial_gausian_median_blur_bilateral_filter.html) is used at the moment.

> [!NOTE]

> There are plenty of similar projects in github. The most of them are implemented in phyton, few in java. ChatGPT can parse

> receipts very well. However, it uses the same tesseract and pretty straight forward parsing rules behind the scene. Also, 

> free version of ChatGPT is limited to processing 2-3 images per day.  

> So, I decided that I can do my own implementation using my favourite Java.

## How to run tesseract via tess4j on MacOS

Install tessract via MAC port

    sudo port install tesseract

In your java application: 

Add VM argument

    

    -Djna.library.path=/opt/local/var/macports/software/tesseract/tesseract-5.4.1_2.darwin_22.x86_64/opt/local/lib

Or add Env variable

    DYLD_LIBRARY_PATH=/opt/local/var/macports/software/tesseract/tesseract-5.4.1_2.darwin_22.x86_64/opt/local/lib

Path to tessdata should be

    /opt/local/share/tessdata/

Maven dependency

        

            net.sourceforge.tess4j

            tess4j

            5.12.0

        

## How to run opencv4 via java

Install opencv4 with java support via MAC port

    sudo port install opencv4 +java

Check the installation

    sudo port contents opencv4 | grep java

If everything is fine tt should show:

    Port opencv4 @4.9.0_4+java contains:

    /opt/local/libexec/opencv4/java/jar/opencv-490.jar

    /opt/local/libexec/opencv4/java/jni/libopencv_java490.dylib

In your java application:

Add VM argument

    -Djava.library.path=/opt/local/libexec/opencv4/java/jni/

Your code must contain the following line before triggering opencv Imgcodecs

    System.loadLibrary(Core.NATIVE_LIBRARY_NAME);

Maven dependency

        

            org.openpnp

            opencv

            4.9.0-0

        

## How to run the pipeline

Build the jar

    mvn clean package

Run java

    java \

        -Djna.library.path=/opt/local/var/macports/software/tesseract/tesseract-5.4.1_2.darwin_22.x86_64/opt/local/lib  \

        -Djava.library.path=/opt/local/libexec/opencv4/java/jni/  \

    -jar target/grocery-expenses-pipeline-1.0-SNAPSHOT.jar  \

        -file src/test/resources/samples/IMG_5213.jpeg  \

        -restEndpoint http://localhost:8080  \

        -storeName PackNSave

where

* -file  -  is a path to an image that needs to be processed. This is a mandatory argument

* -restEndpoint  -  is a REST-endpoint where a result of recognition and parsing needs to be sent as a JSON. This is an optional argument.

* -storeName  -  is a name of store where you got current receipt. By default, it is "Undefined"