hcchappy / ocr-all-in-one

A repository that helps to easily test what OCR is best for your documents

Ocr.Wrapper

This is still in alpha state and we are planning to update some of the API's

It is a wrapper with minimal dependencies that helps you run and compare the most popular OCR services For now we support the following: Windows, Azure Vision, Tesseract, Google OCR, AWs Rekognition

Code example to use with Azure Ocr Service:

[TestMethod()]
public async Task AzureOcr()
{
    var subscriptionKey = ConfigurationManager.AppSettings["azureSubscriptionKey"];
    var endpoint = ConfigurationManager.AppSettings["azureEndpoint"];


    AzureOcrService azureOcrExecutor = new AzureOcrService(subscriptionKey, endpoint);
    AzureOcrResults result = await azureOcrExecutor.GetOcrResultAsync(@"data/abc.JPG");
    Assert.IsNotNull(result);
    GenericOcrResponse genericResult = result.Map();
    Assert.IsNotNull(genericResult);
}

Code example to use Windows OCR:

Tested on Windows 10 and Windows Server 2016 Datacenter version on Azure and it works. It should work also in Windows Server 2019 full.

Not such a known OCR, but from my tests, for a lot of formats it's giving impressing results.

[TestMethod()]
public async Task WindowsOcr()
{
    WindowsOcrService windowsOcrExecutor = new WindowsOcrService();
    WindowsOcrResult result = await windowsOcrExecutor.GetOcrResultAsync(@"data/abc.JPG", "en");
    Assert.IsNotNull(result);
    GenericOcrResponse genericResult = result.Map();
    Assert.IsNotNull(genericResult);
}

Tesseract example: This is one of the most known OCR frameworks, we are basically just a wrapper around the executable.

Tesseract must be installed in order for this to work.

For Windows 10 X64 we have a class: TesseractInstaller that can be used to automatically install Tesseract(just copy binaries)

[TestMethod]
public async Task Tesseract5()
{
    TesseractOcrService tesseractService = new TesseractOcrService();
    TesseractResponse result = await tesseractService.GetOcrResultAsync(@"data/abc.JPG", "eng");
    Assert.IsNotNull(result);
    GenericOcrResponse genericResult = result.Map();
    Assert.IsNotNull(genericResult);
}

Google OCR Service example

[TestMethod]
public async Task GoogleOCR()
{
    var apiToken = ConfigurationManager.AppSettings["googleApiToken"];
    GoogleOcrService googleOcrService = new GoogleOcrService(apiToken);

    GoogleOcrResponse result = await googleOcrService.GetOcrResultAsync(@"data/abc.JPG");
    var descriptions = result.Responses.SelectMany(r => r.Annotations).ToList();
    Assert.IsNotNull(result);
    GenericOcrResponse genericResult = result.Map();
    Assert.IsNotNull(genericResult);
}

AWS Rekognition OCR example

[TestMethod]
public async Task AwsOcr()
{
    var accessKey = ConfigurationManager.AppSettings["awsAccessKey"];
    var secretKey = ConfigurationManager.AppSettings["awsSecretKey"];
    AwsOcrService awsOcrService = new AwsOcrService(accessKey, secretKey);
    AwsOcrResponse result = await awsOcrService.GetOcrResultAsync(@"data/abc.JPG");
    Assert.IsNotNull(result);
    GenericOcrResponse genericResult = result.Map();
    Assert.IsNotNull(genericResult);
}

A way to run all OCRs and get all responses (On settings you can configure Tesseract to be automatically installed)

[TestMethod]
public async Task RunAllOcrs()
{
    StandardOcrSettings standardOcrSettings = GetStandardOcrSettings();

    MultiOcrRunner genericOcrRunner = await new StandardMultiOcrRunnerFactory(standardOcrSettings)
        .GetMultiOcrRunner();
    Dictionary<string, GenericOcrResponse> results = await genericOcrRunner.RunAllOcrEnginesOnImage(@"data/abc.JPG");
    Assert.IsNotNull(results);
}

You can configure file storage to use to cache the results - interface from Foundatio project, so multiple file storage options are available - https://github.com/FoundatioFx/Foundatio#file-storage

[TestMethod]
public async Task RunOcrsWithCache()
{
    StandardOcrSettings standardOcrSettings = GetStandardOcrSettings();

    var fullPath = Path.GetFullPath(@"..\Data\Cache\");
    MultiOcrRunner multiOcrRunner = await new StandardMultiOcrRunnerFactory(standardOcrSettings, fullPath)
        .GetMultiOcrRunner();
    Dictionary<string, GenericOcrResponse> results = await multiOcrRunner.RunAllOcrEnginesOnImage(@"data/abc.JPG");
    Assert.IsNotNull(results);
}

private static StandardOcrSettings GetStandardOcrSettings()
{
    var azureSubscriptionKey = ConfigurationManager.AppSettings["azureSubscriptionKey"];
    var azureEndpoint = ConfigurationManager.AppSettings["azureEndpoint"];
    var googleApiToken = ConfigurationManager.AppSettings["googleApiToken"];
    var awsAcessKey = ConfigurationManager.AppSettings["awsAccessKey"];
    var awsSecretKey = ConfigurationManager.AppSettings["awsSecretKey"];
    StandardOcrSettings standardOcrSettings = new StandardOcrSettings(true)
    {
        AwsOcrSettings = new AwsOcrSettings(awsAcessKey, awsSecretKey),
        AzureOcrSettings = new AzureOcrSettings(azureSubscriptionKey, azureEndpoint),
        GoogleOcrSettings = new GoogleOcrSettings(googleApiToken),
        TesseractOcrSettings = new TesseractOcrSettings(),
        WindowsOcrSettings = new WindowsOcrSettings()
    };
    return standardOcrSettings;
}

Ocr.Wrapper.WebAPI

Web API project configured to run all OCR engines. (Tested with IISExpress/local run/windows docker/linux docker) Default page is the Swagger UI that helps to invoke the API.

In order to configure all Cloud services set the following values in user secrets: "azure:SubscriptionKey", "azure:Endpoint", "google:ApiToken", "aws:AccessKey","aws:SecretKey"