Table of contents

• Android PhotoPay integration instructions
• Quick Start
  • Quick start with the sample app
  • SDK integration
• Device requirements
• PhotoPay SDK integration levels
  • Built-in activities (UISettings)
  • Built-in fragment (RecognizerRunnerFragment)
  • Custom UX with RecognizerRunnerView
  • Direct API
    • Using Direct API for recognition of Android Bitmaps and custom camera frames
    • Using Direct API for String recognition (parsing)
    • Understanding DirectAPI's state machine
    • Using Direct API while RecognizerRunnerView is active
• Available built-in activities and overlays
  • FieldByFieldUISettings and FieldByFieldOverlayController
  • Translation and localization
• Handling processing events with RecognizerRunner and RecognizerRunnerView
• Recognizer concept and RecognizerBundle
  • The Recognizer concept
  • RecognizerBundle
    • Passing Recognizer objects between activities
• List of available recognizers
  • Frame Grabber Recognizer
  • Success Frame Grabber Recognizer
  • BlinkInput recognizer
• Field by field scanning feature
• Processor and Parser
  • The Processor concept
  • List of available processors
    • Parser Group Processor
  • The Parser concept
  • List of available parsers
    • Amount Parser
    • Date Parser
    • IBAN Parser
    • Raw Parser
  • PDF417 recognizer
  • Barcode recognizer
• Embedding PhotoPay inside another SDK
• Processor architecture considerations
  • Reducing the final size of your app
    • Consequences of removing processor architecture
  • Combining PhotoPay with other native libraries
    • Resolving conflict on libc++_shared.so
• Troubleshooting
• FAQ and known issues
• Additional info
  • PhotoPay SDK size
  • API reference
  • Contact

Android PhotoPay integration instructions

The package contains Android Archive (AAR) that contains everything you need to use PhotoPay library. Besides AAR, package also contains a sample project that contains following modules:

• PhotoPay-aMinimalSample demonstrates quick and simple integration of PhotoPay library
• PhotoPay-AllRecognizersSample demonstrates integration of almost all available features. This sample application is best for performing a quick test of supported features.
• PhotoPay-CustomFieldByFieldScanSample demonstrates advanced integration of Field by Field feature within custom scan activity. It shows how to create a custom scan activity for scanning little text fields.
• PhotoPay-CustomUISample demonstrates advanced integration within custom scan activity.
• PhotoPay-DirectApiSample demonstrates how to perform scanning of Android Bitmaps

The source code of all sample apps is given to you to show you how to perform integration of PhotoPay SDK into your app. You can use this source code and all resources as you wish. You can use sample apps as a basis for creating your own app, or you can copy/paste the code and/or resources from sample apps into your app and use them as you wish without even asking us for permission.

PhotoPay is supported on Android SDK version 21 (Android 5.0) or later.

The list of all provided scan activities can be found in the Available built-in activities and overlays section.

You can also create your own scanning UI - you just need to embed RecognizerRunnerView into your activity and pass activity's lifecycle events to it and it will control the camera and recognition process. For more information, see Embedding RecognizerRunnerView into custom scan activity.

Quick Start

Quick start with the sample app

1. Open Android Studio.
2. In Quick Start dialog choose Import project (Eclipse ADT, Gradle, etc.).
3. In File dialog select PhotopaySample folder.
4. Wait for the project to load. If Android studio asks you to reload project on startup, select Yes.

SDK integration

Adding PhotoPay dependency

1. Create a libs folder in your Android Studio project.

2. Download the LibPhotoPay.aar file and move it to the created libs folder.

3. In your app's build.gradle, add a dependency to LibPhotoPay (by using a relative path to the LibPhotoPay.aar file) and appcompat:

   dependencies {
       implementation files('../libs/LibPhotoPay.aar')
       implementation "androidx.appcompat:appcompat:1.6.1"
   }
   

Importing Javadoc

1. In Android Studio project sidebar, ensure project view is enabled
2. Expand External Libraries entry (usually this is the last entry in project view)
3. Locate LibPhotoPay-unspecified entry, right click on it and select Library Properties...
4. A Library Properties pop-up window will appear
5. Click the + button in bottom left corner of the window
6. Window for choosing JAR file will appear
7. Find and select LibPhotoPay-javadoc.jar file which is located in root folder of the SDK distribution
8. Click OK

Performing your first scan

1. A valid license key is required to initialize scanning. You can request a free trial license key, after you register, at Microblink Developer Hub. License is bound to package name of your app, so please make sure you enter the correct package name when asked.

   Download your licence file and put it in your application's assets folder. Make sure to set the license key before using any other classes from the SDK, otherwise you will get a runtime exception.

   We recommend that you extend Android Application class and set the license in onCreate callback like this:

   public class MyApplication extends Application {
       @Override
       public void onCreate() {
           MicroblinkSDK.setLicenseFile("path/to/license/file/within/assets/dir", this);
       }
   }

2. In your main activity, create recognizer objects that will perform image recognition, configure them and put them into RecognizerBundle object. You can see more information about available recognizers and RecognizerBundle here.

   For example, to scan Croatian slip, configure your recognizer like this:

   public class MyActivity extends Activity {
       private CroatiaSlipRecognizer mRecognizer;
       private RecognizerBundle mRecognizerBundle;
       
       @Override
       protected void onCreate(Bundle bundle) {
           super.onCreate(bundle);
           
           // setup views, as you would normally do in onCreate callback
           
           // create CroatiaSlipRecognizer
           mRecognizer = new CroatiaSlipRecognizer();
           
           // bundle recognizers into RecognizerBundle
           mRecognizerBundle = new RecognizerBundle(mRecognizer);
       }
   }

3. Start recognition process by creating PhotopayUISettings and calling ActivityRunner.startActivityForResult:

   // method within MyActivity from previous step
   public void startScanning() {
       // Settings for PhotopayActivity
       PhotopayUISettings settings = new PhotopayUISettings(mRecognizerBundle);
       
       // tweak settings as you wish
       
       // Start activity
       ActivityRunner.startActivityForResult(this, MY_REQUEST_CODE, settings);
   }

4. onActivityResult will be called in your activity after scanning is finished, here you can get the scanning results.

   @Override
   protected void onActivityResult(int requestCode, int resultCode, Intent data) {
       super.onActivityResult(requestCode, resultCode, data);
       
       if (requestCode == MY_REQUEST_CODE) {
           if (resultCode == Activity.RESULT_OK && data != null) {
               // load the data into all recognizers bundled within your RecognizerBundle
               mRecognizerBundle.loadFromIntent(data);
               
               // now every recognizer object that was bundled within RecognizerBundle
               // has been updated with results obtained during scanning session
               
               // you can get the result by invoking getResult on recognizer
               CroatiaSlipRecognizer.Result result = mRecognizer.getResult();
               if (result.getResultState() == Recognizer.Result.State.Valid) {
                   // result is valid, you can use it however you wish
               }
           }
       }
   }

   For more information about available recognizers and RecognizerBundle, see RecognizerBundle and available recognizers.

Device requirements

Android Version

PhotoPay requires Android API level 21 or newer.

Camera

Camera video preview resolution also matters. In order to perform successful scans, camera preview resolution must be at least 720p. Note that camera preview resolution is not the same as video recording resolution.

Processor architecture

PhotoPay is distributed with ARMv7 and ARM64 native library binaries.

PhotoPay is a native library, written in C++ and available for multiple platforms. Because of this, PhotoPay cannot work on devices with obscure hardware architectures. We have compiled PhotoPay native code only for the most popular Android ABIs.

Even before setting the license key, you should check if the PhotoPay is supported on the current device (see next section: Compatibility check). Attempting to call any method from the SDK that relies on native code, such as license check, on a device with unsupported CPU architecture will crash your app.

If you are combining PhotoPay library with other libraries that contain native code into your application, make sure you match the architectures of all native libraries.

For more information, see Processor architecture considerations section.

Compatibility check

Here's how you can check whether the PhotoPay is supported on the device:

Java

// check if PhotoPay is supported on the device,
RecognizerCompatibilityStatus status = RecognizerCompatibility.getRecognizerCompatibilityStatus(this);
if (status == RecognizerCompatibilityStatus.RECOGNIZER_SUPPORTED) {
    Toast.makeText(this, "PhotoPay is supported!", Toast.LENGTH_LONG).show();
} else if (status == RecognizerCompatibilityStatus.NO_CAMERA) {
    Toast.makeText(this, "PhotoPay is supported only via Direct API!", Toast.LENGTH_LONG).show();
} else if (status == RecognizerCompatibilityStatus.PROCESSOR_ARCHITECTURE_NOT_SUPPORTED) {
    Toast.makeText(this, "PhotoPay is not supported on current processor architecture!", Toast.LENGTH_LONG).show();
} else {
    Toast.makeText(this, "PhotoPay is not supported! Reason: " + status.name(), Toast.LENGTH_LONG).show();
}

Kotlin

// check if _PhotoPay_ is supported on the device,
when (val status = RecognizerCompatibility.getRecognizerCompatibilityStatus(this)) {
    RecognizerCompatibilityStatus.RECOGNIZER_SUPPORTED -> {
        Toast.makeText(this, "PhotoPay is supported!", Toast.LENGTH_LONG).show()
    }
    RecognizerCompatibilityStatus.NO_CAMERA -> {
        Toast.makeText(this, "PhotoPay is supported only via Direct API!", Toast.LENGTH_LONG).show()
    }
    RecognizerCompatibilityStatus.PROCESSOR_ARCHITECTURE_NOT_SUPPORTED -> {
        Toast.makeText(this, "PhotoPay is not supported on current processor architecture!", Toast.LENGTH_LONG).show()
    }
    else -> {
        Toast.makeText(this, "PhotoPay is not supported! Reason: " + status.name, Toast.LENGTH_LONG).show()
    }
}

Some recognizers require camera with autofocus. If you try using them on a device that doesn't support autofocus, you will get an error. To prevent that, you can check whether a recognizer requires autofocus by calling its requiresAutofocus method.

If you already have an array of recognizers, you can easily filter out recognizers that require autofocus from array using the following code snippet:

Java

Recognizer[] recArray = ...;
if(!RecognizerCompatibility.cameraHasAutofocus(CameraType.CAMERA_BACKFACE, this)) {
    recArray = RecognizerUtils.filterOutRecognizersThatRequireAutofocus(recArray);
}

Kotlin

var recArray: Array<Recognizer> = ...
if(!RecognizerCompatibility.cameraHasAutofocus(CameraType.CAMERA_BACKFACE, this)) {
    recArray = RecognizerUtils.filterOutRecognizersThatRequireAutofocus(recArray)
}

PhotoPay SDK integration levels

You can integrate PhotoPay into your app in four different ways, depending on your use case and customisation needs:

1. Built-in activities (UISettings) - SDK handles everything and you just need to start our built-in activity and handle result, customisation options are limited
2. Built-in fragment (RecognizerRunnerFragment) - reuse scanning UX from our built-in activities in your own activity
3. Custom UX (RecognizerRunnerView) - SDK handles camera management while you have to implement completely custom scanning UX
4. Direct Api (RecognizerRunner) - SKD only handles recognition while you have to provide it with the images, either from camera or from a file

Built-in activities (UISettings)

UISettings is a class that contains all the necessary settings for SDK's built-in scan activities. It configures scanning activity behaviour, strings, icons and other UI elements. As shown in the first scan example, you should use ActivityRunner to start the scan activity configured by UISettings.

We provide multiple UISettings classes specialised for different scanning scenarios. Each UISettings object has properties which can be changed via appropriate setter methods. For example, you can customise camera settings with setCameraSettings metod.

All available UISettings classes are listed here.

Built-in fragment (RecognizerRunnerFragment)

If you want to reuse our built-in activity UX inside your own activity, use RecognizerRunnerFragment. Activity that will host RecognizerRunnerFragment must implement ScanningOverlayBinder interface. Attempting to add RecognizerRunnerFragment to activity that does not implement that interface will result in ClassCastException.

The ScanningOverlayBinder is responsible for returning non-null implementation of ScanningOverlay - class that will manage UI on top of RecognizerRunnerFragment. It is not recommended to create your own ScanningOverlay implementation, use one of our implementations listed here instead.

Here is the minimum example for activity that hosts the RecognizerRunnerFragment:

Java

public class MyActivity extends AppCompatActivity implements RecognizerRunnerFragment.ScanningOverlayBinder {
    private CroatiaSlipRecognizer mRecognizer;
    private RecognizerBundle mRecognizerBundle;
    private BasicOverlayController mScanOverlay;
    private RecognizerRunnerFragment mRecognizerRunnerFragment;

    @Override
    protected void onCreate(Bundle savedInstanceState) {
        super.onCreate();
        setContentView(R.layout.activity_my_activity);
        mScanOverlay = createOverlay();
        if (null == savedInstanceState) {
            // create fragment transaction to replace R.id.recognizer_runner_view_container with RecognizerRunnerFragment
            mRecognizerRunnerFragment = new RecognizerRunnerFragment();
            FragmentTransaction fragmentTransaction = getSupportFragmentManager().beginTransaction();
            fragmentTransaction.replace(R.id.recognizer_runner_view_container, mRecognizerRunnerFragment);
            fragmentTransaction.commit();
        } else {
            // obtain reference to fragment restored by Android within super.onCreate() call
            mRecognizerRunnerFragment = (RecognizerRunnerFragment) getSupportFragmentManager().findFragmentById(R.id.recognizer_runner_view_container);
        }
    }

    @Override
    @NonNull
    public ScanningOverlay getScanningOverlay() {
        return mScanOverlay;
    }

    private BasicOverlayController createOverlay() {
        // create CroatiaSlipRecognizer
        mRecognizer = new CroatiaSlipRecognizer();

        // bundle recognizers into RecognizerBundle
        mRecognizerBundle = new RecognizerBundle(mRecognizer);

        PhotopayUISettings settings = new PhotopayUISettings(mRecognizerBundle);

        return settings.createOverlayController(this, mScanResultListener);
    }

    private final ScanResultListener mScanResultListener = new ScanResultListener() {
        @Override
        public void onScanningDone(@NonNull RecognitionSuccessType recognitionSuccessType) {
            // pause scanning to prevent new results while fragment is being removed
            mRecognizerRunnerFragment.getRecognizerRunnerView().pauseScanning();

            // now you can remove the RecognizerRunnerFragment with new fragment transaction
            // and use result within mRecognizer safely without the need for making a copy of it

            // if not paused, as soon as this method ends, RecognizerRunnerFragments continues
            // scanning. Note that this can happen even if you created fragment transaction for
            // removal of RecognizerRunnerFragment - in the time between end of this method
            // and beginning of execution of the transaction. So to ensure result within mRecognizer
            // does not get mutated, ensure calling pauseScanning() as shown above.
        }
        @Override
        public void onUnrecoverableError(@NonNull Throwable throwable) {
        }
    };
    
}

Kotlin Compose

package com.microblink.blinkid

class MainActivity : AppCompatActivity(), RecognizerRunnerFragment.ScanningOverlayBinder {
    private lateinit var mRecognizer: CroatiaSlipRecognizer
    private lateinit var mRecognizerRunnerFragment: RecognizerRunnerFragment
    private lateinit var mRecognizerBundle: RecognizerBundle
    private lateinit var mScanOverlay: BasicOverlayController

    override fun onCreate(savedInstanceState: Bundle?) {
        super.onCreate(savedInstanceState)
        if (!::mScanOverlay.isInitialized) {
            mScanOverlay = createOverlayController()
        }
        setContent {
            this.run {
                // viewBinding has to be set to 'true' in buildFeatures block of the build.gradle file
                AndroidViewBinding(RecognizerRunnerLayoutBinding::inflate) {
                    mRecognizerRunnerFragment =
                        fragmentContainerView.getFragment<RecognizerRunnerFragment>()
                }
            }
        }
    }

    override fun getScanningOverlay(): ScanningOverlay {
        return mScanOverlay
    }

    private fun createOverlay(): BasicOverlayController {
        // create CroatiaSlipRecognizer
        val mRecognizer = CroatiaSlipRecognizer()

        // bundle recognizers into RecognizerBundle
        mRecognizerBundle = RecognizerBundle(mRecognizer)

        val settings = PhotopayUISettings(mRecognizerBundle)

        return settings.createOverlayController(this, mScanResultListener)
    }

    private val mScanResultListener: ScanResultListener = object : ScanResultListener {
        override fun onScanningDone(p0: RecognitionSuccessType) {
            // pause scanning to prevent new results while fragment is being removed
            mRecognizerRunnerFragment!!.recognizerRunnerView!!.pauseScanning()

            // now you can remove the RecognizerRunnerFragment with new fragment transaction
            // and use result within mRecognizer safely without the need for making a copy of it

            // if not paused, as soon as this method ends, RecognizerRunnerFragments continues
            // scanning. Note that this can happen even if you created fragment transaction for
            // removal of RecognizerRunnerFragment - in the time between end of this method
            // and beginning of execution of the transaction. So to ensure result within mRecognizer
            // does not get mutated, ensure calling pauseScanning() as shown above.
        }
        override fun onUnrecoverableError(p0: Throwable) {
        }
    }
    
}

Please refer to sample apps provided with the SDK for more detailed example and make sure your host activity's orientation is set to nosensor or has configuration changing enabled (i.e. is not restarted when configuration change happens). For more information, check scan orientation section.

Custom UX with RecognizerRunnerView

This section discusses how to embed RecognizerRunnerView into your scan activity and perform scan.

1. First make sure that RecognizerRunnerView is a member field in your activity. This is required because you will need to pass all activity's lifecycle events to RecognizerRunnerView.
2. It is recommended to keep your scan activity in one orientation, such as portrait or landscape. Setting sensor as scan activity's orientation will trigger full restart of activity whenever device orientation changes. This will provide very poor user experience because both camera and PhotoPay native library will have to be restarted every time. There are measures against this behaviour that are discussed later.
3. In your activity's onCreate method, create a new RecognizerRunnerView, set RecognizerBundle containing recognizers that will be used by the view, define CameraEventsListener that will handle mandatory camera events, define ScanResultListener that will receive call when recognition has been completed and then call its create method. After that, add your views that should be layouted on top of camera view.
4. Pass in your activity's lifecycle using setLifecycle method to enable automatic handling of lifeceycle events.

Here is the minimum example of integration of RecognizerRunnerView as the only view in your activity:

public class MyScanActivity extends AppCompatActivity {
    private static final int PERMISSION_CAMERA_REQUEST_CODE = 42;
    private RecognizerRunnerView mRecognizerRunnerView;
    private CroatiaSlipRecognizer mRecognizer;
    private RecognizerBundle mRecognizerBundle;

    @Override
    protected void onCreate(Bundle savedInstanceState) {
        super.onCreate(savedInstanceState);

        // create CroatiaSlipRecognizer
        mRecognizer = new CroatiaSlipRecognizer();

        // bundle recognizers into RecognizerBundle
        mRecognizerBundle = new RecognizerBundle(mRecognizer);
        // create RecognizerRunnerView
        mRecognizerRunnerView = new RecognizerRunnerView(this);
        
        // set lifecycle to automatically call recognizer runner view lifecycle methods
        mRecognizerRunnerView.setLifecycle(getLifecycle());

        // associate RecognizerBundle with RecognizerRunnerView
        mRecognizerRunnerView.setRecognizerBundle(mRecognizerBundle);

        // scan result listener will be notified when scanning is complete
        mRecognizerRunnerView.setScanResultListener(mScanResultListener);
        // camera events listener will be notified about camera lifecycle and errors
        mRecognizerRunnerView.setCameraEventsListener(mCameraEventsListener);

        setContentView(mRecognizerRunnerView);
    }

    @Override
    public void onConfigurationChanged(Configuration newConfig) {
        super.onConfigurationChanged(newConfig);
        // changeConfiguration is not handled by lifecycle events so call it manually
        mRecognizerRunnerView.changeConfiguration(newConfig);
    }

    private final CameraEventsListener mCameraEventsListener = new CameraEventsListener() {
        @Override
        public void onCameraPreviewStarted() {
            // this method is from CameraEventsListener and will be called when camera preview starts
        }

        @Override
        public void onCameraPreviewStopped() {
            // this method is from CameraEventsListener and will be called when camera preview stops
        }

        @Override
        public void onError(Throwable exc) {
            /**
             * This method is from CameraEventsListener and will be called when
             * opening of camera resulted in exception or recognition process
             * encountered an error. The error details will be given in exc
             * parameter.
             */
        }

        @Override
        @TargetApi(23)
        public void onCameraPermissionDenied() {
            /**
             * Called in Android 6.0 and newer if camera permission is not given
             * by user. You should request permission from user to access camera.
             */
            requestPermissions(new String[]{Manifest.permission.CAMERA}, PERMISSION_CAMERA_REQUEST_CODE);
            /**
             * Please note that user might have not given permission to use
             * camera. In that case, you have to explain to user that without
             * camera permissions scanning will not work.
             * For more information about requesting permissions at runtime, check
             * this article:
             * https://developer.android.com/training/permissions/requesting.html
             */
        }

        @Override
        public void onAutofocusFailed() {
            /**
             * This method is from CameraEventsListener will be called when camera focusing has failed.
             * Camera manager usually tries different focusing strategies and this method is called when all
             * those strategies fail to indicate that either object on which camera is being focused is too
             * close or ambient light conditions are poor.
             */
        }

        @Override
        public void onAutofocusStarted(Rect[] areas) {
            /**
             * This method is from CameraEventsListener and will be called when camera focusing has started.
             * You can utilize this method to draw focusing animation on UI.
             * Areas parameter is array of rectangles where focus is being measured.
             * It can be null on devices that do not support fine-grained camera control.
             */
        }

        @Override
        public void onAutofocusStopped(Rect[] areas) {
            /**
             * This method is from CameraEventsListener and will be called when camera focusing has stopped.
             * You can utilize this method to remove focusing animation on UI.
             * Areas parameter is array of rectangles where focus is being measured.
             * It can be null on devices that do not support fine-grained camera control.
             */
        }
    };
    
    private final ScanResultListener mScanResultListener = new ScanResultListener() {
        @Override
        public void onScanningDone(@NonNull RecognitionSuccessType recognitionSuccessType) {
            // this method is from ScanResultListener and will be called when scanning completes
            // you can obtain scanning result by calling getResult on each
            // recognizer that you bundled into RecognizerBundle.
            // for example:

            CroatiaSlipRecognizer.Result result = mRecognizer.getResult();
            if (result.getResultState() == Recognizer.Result.State.Valid) {
                // result is valid, you can use it however you wish
            }

            // Note that mRecognizer is stateful object and that as soon as
            // scanning either resumes or its state is reset
            // the result object within mRecognizer will be changed. If you
            // need to create a immutable copy of the result, you can do that
            // by calling clone() on it, for example:

            CroatiaSlipRecognizer.Result immutableCopy = result.clone();

            // After this method ends, scanning will be resumed and recognition
            // state will be retained. If you want to prevent that, then
            // you should call:
            mRecognizerRunnerView.resetRecognitionState();
            // Note that reseting recognition state will clear internal result
            // objects of all recognizers that are bundled in RecognizerBundle
            // associated with RecognizerRunnerView.

            // If you want to pause scanning to prevent receiving recognition
            // results or mutating result, you should call:
            mRecognizerRunnerView.pauseScanning();
            // if scanning is paused at the end of this method, it is guaranteed
            // that result within mRecognizer will not be mutated, therefore you
            // can avoid creating a copy as described above

            // After scanning is paused, you will have to resume it with:
            mRecognizerRunnerView.resumeScanning(true);
            // boolean in resumeScanning method indicates whether recognition
            // state should be automatically reset when resuming scanning - this
            // includes clearing result of mRecognizer
        }
    };
    
}

Scan activity's orientation

If activity's screenOrientation property in AndroidManifest.xml is set to sensor, fullSensor or similar, activity will be restarted every time device changes orientation from portrait to landscape and vice versa. While restarting activity, its onPause, onStop and onDestroy methods will be called and then new activity will be created anew. This is a potential problem for scan activity because in its lifecycle it controls both camera and native library - restarting the activity will trigger both restart of the camera and native library. This is a problem because changing orientation from landscape to portrait and vice versa will be very slow, thus degrading a user experience. We do not recommend such setting.

For that matter, we recommend setting your scan activity to either portrait or landscape mode and handle device orientation changes manually. To help you with this, RecognizerRunnerView supports adding child views to it that will be rotated regardless of activity's screenOrientation. You add a view you wish to be rotated (such as view that contains buttons, status messages, etc.) to RecognizerRunnerView with addChildView method. The second parameter of the method is a boolean that defines whether the view you are adding will be rotated with device. To define allowed orientations, implement OrientationAllowedListener interface and add it to RecognizerRunnerView with method setOrientationAllowedListener. This is the recommended way of rotating camera overlay.

However, if you really want to set screenOrientation property to sensor or similar and want Android to handle orientation changes of your scan activity, then we recommend to set configChanges property of your activity to orientation|screenSize. This will tell Android not to restart your activity when device orientation changes. Instead, activity's onConfigurationChanged method will be called so that activity can be notified of the configuration change. In your implementation of this method, you should call changeConfiguration method of RecognizerView so it can adapt its camera surface and child views to new configuration.

Direct API

This section will describe how to use direct API to recognize android Bitmaps without the need for camera. You can use direct API anywhere from your application, not just from activities.

Image recognition performance highly depends on the quality of the input images. When our camera management is used (scanning from a camera), we do our best to get camera frames with the best possible quality for the used device. On the other hand, when Direct API is used, you need to provide high-quality images without blur and glare for successful recognition.

Using Direct API for recognition of Android Bitmaps and custom camera frames

1. First, you need to obtain reference to RecognizerRunner singleton using getSingletonInstance.
2. Second, you need to initialize the recognizer runner.
3. After initialization, you can use singleton to process:

• Still Android Bitmaps obtained, for example, from the gallery. Use recognizeBitmap or recognizeBitmapWithRecognizers.
• Video Images that are built from custom camera video frames, for example, when you use your own or third party camera management. Recognition will be optimized for speed and will rely on time-redundancy between consecutive video frames in order to yield best possible recognition result. Use recognizeVideoImage or recognizeVideoImageWithRecognizers.
• Still Images when you need thorough scanning of single or few images which are not part of the video stream and you want to get best possible results from the single InputImage. InputImage type comes from our SDK or it can be created by using ImageBuilder. Use recognizeStillImage or recognizeStillImageWithRecognizers.

4. When you want to delete all cached data from multiple recognitions, for example when you want to scan other document and/or restart scanning, you need to reset the recognition state.
5. Do not forget to terminate the recognizer runner singleton after usage (it is a shared resource).

Here is the minimum example of usage of direct API for recognizing android Bitmap:

public class DirectAPIActivity extends Activity {
    private RecognizerRunner mRecognizerRunner;
    private CroatiaSlipRecognizer mRecognizer;
    private RecognizerBundle mRecognizerBundle;

    @Override
    protected void onCreate(Bundle savedInstanceState) {
        super.onCreate();
        // initialize your activity here
        // create CroatiaSlipRecognizer
        mRecognizer = new CroatiaSlipRecognizer();

        // bundle recognizers into RecognizerBundle
        mRecognizerBundle = new RecognizerBundle(mRecognizer);

        try {
            mRecognizerRunner = RecognizerRunner.getSingletonInstance();
        } catch (FeatureNotSupportedException e) {
            Toast.makeText(this, "Feature not supported! Reason: " + e.getReason().getDescription(), Toast.LENGTH_LONG).show();
            finish();
            return;
        }

        mRecognizerRunner.initialize(this, mRecognizerBundle, new DirectApiErrorListener() {
            @Override
            public void onRecognizerError(Throwable t) {
                Toast.makeText(DirectAPIActivity.this, "There was an error in initialization of Recognizer: " + t.getMessage(), Toast.LENGTH_SHORT).show();
                finish();
            }
        });
    }

    @Override
    protected void onResume() {
        super.onResume();
        // start recognition
        Bitmap bitmap = BitmapFactory.decodeFile("/path/to/some/file.jpg");
        mRecognizerRunner.recognizeBitmap(bitmap, Orientation.ORIENTATION_LANDSCAPE_RIGHT, mScanResultListener);
    }

    @Override
    protected void onDestroy() {
        super.onDestroy();
        mRecognizerRunner.terminate();
    }

    private final ScanResultListener mScanResultListener = new ScanResultListener() {
        @Override
        public void onScanningDone(@NonNull RecognitionSuccessType recognitionSuccessType) {
            // this method is from ScanResultListener and will be called
            // when scanning completes
            // you can obtain scanning result by calling getResult on each
            // recognizer that you bundled into RecognizerBundle.
            // for example:

            CroatiaSlipRecognizer.Result result = mRecognizer.getResult();
            if (result.getResultState() == Recognizer.Result.State.Valid) {
                // result is valid, you can use it however you wish
            }
        }
    };

}

ScanResultListener.onScanningDone method is called for each input image that you send to the recognition. You can call RecognizerRunner.recognize* method multiple times with different images of the same document for better reading accuracy until you get a successful result in the listener's onScanningDone method. This is useful when you are using your own or third-party camera management.

Using Direct API for String recognition (parsing)

Some recognizers support recognition from String. They can be used through Direct API to parse given String and return data just like when they are used on an input image. When recognition is performed on String, there is no need for the OCR. Input String is used in the same way as the OCR output is used when image is being recognized.

Recognition from String can be performed in the same way as recognition from image, described in the previous section.

The only difference is that one of the RecognizerRunner singleton methods for recognition from string should be called:

• recognizeString
• recognizeStringWithRecognizers

Understanding DirectAPI's state machine

Direct API's RecognizerRunner singleton is a state machine that can be in one of 3 states: OFFLINE, READY and WORKING.

• When you obtain the reference to RecognizerRunner singleton, it will be in OFFLINE state.
• You can initialize RecognizerRunner by calling initialize method. If you call initialize method while RecognizerRunner is not in OFFLINE state, you will get IllegalStateException.
• After successful initialization, RecognizerRunner will move to READY state. Now you can call any of the recognize* methods.
• When starting recognition with any of the recognize* methods, RecognizerRunner will move to WORKING state. If you attempt to call these methods while RecognizerRunner is not in READY state, you will get IllegalStateException
• Recognition is performed on background thread so it is safe to call all RecognizerRunner's methods from UI thread
• When recognition is finished, RecognizerRunner first moves back to READY state and then calls the onScanningDone method of the provided ScanResultListener.
• Please note that ScanResultListener's onScanningDone method will be called on background processing thread, so make sure you do not perform UI operations in this callback. Also note that until the onScanningDone method completes, RecognizerRunner will not perform recognition of another image or string, even if any of the recognize* methods have been called just after transitioning to READY state. This is to ensure that results of the recognizers bundled within RecognizerBundle associated with RecognizerRunner are not modified while possibly being used within onScanningDone method.
• By calling terminate method, RecognizerRunner singleton will release all its internal resources. Note that even after calling terminate you might receive onScanningDone event if there was work in progress when terminate was called.
• terminate method can be called from any RecognizerRunner singleton's state
• You can observe RecognizerRunner singleton's state with method getCurrentState

Using Direct API while RecognizerRunnerView is active

Both RecognizerRunnerView and RecognizerRunner use the same internal singleton that manages native code. This singleton handles initialization and termination of native library and propagating recognizers to native library. It is possible to use RecognizerRunnerView and RecognizerRunner together, as internal singleton will make sure correct synchronization and correct recognition settings are used. If you run into problems while using RecognizerRunner in combination with RecognizerRunnerView, let us know!

Available built-in activities and overlays

PhotopayUISettings

PhotopayUISettings launches activity that uses BasicOverlayController with UI best suited for scanning various payment slips.

OcrLineUISettings

OcrLineUISettings launches activity that uses BasicOverlayController with UI best suited for performing scanning of payment slips that have entire payment information encoded in OCR line in lower part of the slip. For example, payment slips in Kosovo, Netherlands, Switzerland and United Kingdom.

BarcodeUISettings

BarcodeUISettings launches activity that uses BasicOverlayController with UI best suited for performing scanning of various barcodes.

FieldByFieldUISettings and FieldByFieldOverlayController

FieldByFieldOverlayController is best suited for performing scanning of small text fields, which are scanned in the predefined order, one by one.

To launch a built-in activity that uses FieldByFieldOverlayController use FieldByFieldUISettings.

Translation and localization

Strings used within built-in activities and overlays can be localized to any language. If you are using RecognizerRunnerView (see this chapter for more information) in your custom scan activity or fragment, you should handle localization as in any other Android app. RecognizerRunnerView does not use strings nor drawables, it only uses assets from assets/microblink folder. Those assets must not be touched as they are required for recognition to work correctly.

However, if you use our built-in activities or overlays, they will use resources packed within LibPhotoPay.aar to display strings and images on top of the camera view. We have already prepared strings for several languages which you can use out of the box. You can also modify those strings, or you can add your own language.

To use a language, you have to enable it from the code:

• To use a certain language, on application startup, before opening any UI component from the SDK, you should call method LanguageUtils.setLanguageAndCountry(language, country, context). For example, you can set language to Croatian like this:

  // define PhotoPay language
  LanguageUtils.setLanguageAndCountry("hr", "", this);

Adding new language

PhotoPay can easily be translated to other languages. The res folder in LibPhotoPay.aar archive has folder values which contains strings.xml - this file contains english strings. In order to make e.g. croatian translation, create a folder values-hr in your project and put the copy of strings.xml inside it (you might need to extract LibPhotoPay.aar archive to access those files). Then, open that file and translate the strings from English into Croatian.

Changing strings in the existing language

To modify an existing string, the best approach would be to:

1. Choose a language you want to modify. For example Croatian ('hr').
2. Find strings.xml in folder res/values-hr of the LibPhotoPay.aar archive
3. Choose a string key which you want to change. For example: <string name="MBBack">Back</string>
4. In your project create a file strings.xml in the folder res/values-hr, if it doesn't already exist
5. Create an entry in the file with the value for the string which you want. For example: <string name="MBBack">Natrag</string>
6. Repeat for all the string you wish to change

Handling processing events with RecognizerRunner and RecognizerRunnerView

Processing events, also known as Metadata callbacks are purely intended for giving processing feedback on UI or to capture some debug information during development of your app using PhotoPay SDK. For that reason, built-in activities and fragments handle those events internally. If you need to handle those events yourself, you need to use either RecognizerRunnerView or RecognizerRunner.

Callbacks for all events are bundled into the MetadataCallbacks object. Both RecognizerRunner and RecognizerRunnerView have methods which allow you to set all your callbacks.

We suggest that you check for more information about available callbacks and events to which you can handle in the javadoc for MetadataCallbacks class.

Please note that both those methods need to pass information about available callbacks to the native code and for efficiency reasons this is done at the time setMetadataCallbacks method is called and not every time when change occurs within the MetadataCallbacks object. This means that if you, for example, set QuadDetectionCallback to MetadataCallbacks after you already called setMetadataCallbacks method, the QuadDetectionCallback will not be registered with the native code and you will not receive its events.

Similarly, if you, for example, remove the QuadDetectionCallback from MetadataCallbacks object after you already called setMetadataCallbacks method, your app will crash with NullPointerException when our processing code attempts to invoke the method on removed callback (which is now set to null). We deliberately do not perform null check here because of two reasons:

• it is inefficient
• having null callback, while still being registered to native code is illegal state of your program and it should therefore crash

Remember, each time you make some changes to MetadataCallbacks object, you need to apply those changes to to your RecognizerRunner or RecognizerRunnerView by calling its setMetadataCallbacks method.

Recognizer concept and RecognizerBundle

This section will first describe what is a Recognizer and how it should be used to perform recognition of the images, videos and camera stream. Next, we will describe how RecognizerBundle can be used to tweak the recognition procedure and to transfer Recognizer objects between activities.

RecognizerBundle is an object which wraps the Recognizers and defines settings about how recognition should be performed. Besides that, RecognizerBundle makes it possible to transfer Recognizer objects between different activities, which is required when using built-in activities to perform scanning, as described in first scan section, but is also handy when you need to pass Recognizer objects between your activities.

List of all available Recognizer objects, with a brief description of each Recognizer, its purpose and recommendations how it should be used to get best performance and user experience, can be found here .

The Recognizer concept

The Recognizer is the basic unit of processing within the PhotoPay SDK. Its main purpose is to process the image and extract meaningful information from it. As you will see later, the PhotoPay SDK has lots of different Recognizer objects that have various purposes.

Each Recognizer has a Result object, which contains the data that was extracted from the image. The Result object is a member of corresponding Recognizer object and its lifetime is bound to the lifetime of its parent Recognizer object. If you need your Result object to outlive its parent Recognizer object, you must make a copy of it by calling its method clone().

Every Recognizer is a stateful object, that can be in two states: idle state and working state. While in idle state, you can tweak Recognizer object's properties via its getters and setters. After you bundle it into a RecognizerBundle and use either RecognizerRunner or RecognizerRunnerView to run the processing with all Recognizer objects bundled within RecognizerBundle, it will change to working state where the Recognizer object is being used for processing. While being in working state, you cannot tweak Recognizer object's properties. If you need to, you have to create a copy of the Recognizer object by calling its clone(), then tweak that copy, bundle it into a new RecognizerBundle and use reconfigureRecognizers to ensure new bundle gets used on processing thread.

While Recognizer object works, it changes its internal state and its result. The Recognizer object's Result always starts in Empty state. When corresponding Recognizer object performs the recognition of given image, its Result can either stay in Empty state (in case Recognizer failed to perform recognition), move to Uncertain state (in case Recognizer performed the recognition, but not all mandatory information was extracted), move to StageValid state (in case Recognizer successfully scanned one part/side of the document and there are more fields to extract) or move to Valid state (in case Recognizer performed recognition and all mandatory information was successfully extracted from the image).

As soon as one Recognizer object's Result within RecognizerBundle given to RecognizerRunner or RecognizerRunnerView changes to Valid state, the onScanningDone callback will be invoked on same thread that performs the background processing and you will have the opportunity to inspect each of your Recognizer objects' Results to see which one has moved to Valid state.

As already stated in section about RecognizerRunnerView, as soon as onScanningDone method ends, the RecognizerRunnerView will continue processing new camera frames with same Recognizer objects, unless paused. Continuation of processing or resetting recognition will modify or reset all Recognizer objects's Results. When using built-in activities, as soon as onScanningDone is invoked, built-in activity pauses the RecognizerRunnerView and starts finishing the activity, while saving the RecognizerBundle with active Recognizer objects into Intent so they can be transferred back to the calling activities.

RecognizerBundle

The RecognizerBundle is wrapper around Recognizers objects that can be used to transfer Recognizer objects between activities and to give Recognizer objects to RecognizerRunner or RecognizerRunnerView for processing.

The RecognizerBundle is always constructed with array of Recognizer objects that need to be prepared for recognition (i.e. their properties must be tweaked already). The varargs constructor makes it easier to pass Recognizer objects to it, without the need of creating a temporary array.

The RecognizerBundle manages a chain of Recognizer objects within the recognition process. When a new image arrives, it is processed by the first Recognizer in chain, then by the second and so on, iterating until a Recognizer object's Result changes its state to Valid or all of the Recognizer objects in chain were invoked (none getting a Valid result state). If you want to invoke all Recognizers in the chain, regardless of whether some Recognizer object's Result in chain has changed its state to Valid or not, you can allow returning of multiple results on a single image.

You cannot change the order of the Recognizer objects within the chain - no matter the order in which you give Recognizer objects to RecognizerBundle, they are internally ordered in a way that provides best possible performance and accuracy. Also, in order for PhotoPay SDK to be able to order Recognizer objects in recognition chain in the best way possible, it is not allowed to have multiple instances of Recognizer objects of the same type within the chain. Attempting to do so will crash your application.

Passing Recognizer objects between activities

Besides managing the chain of Recognizer objects, RecognizerBundle also manages transferring bundled Recognizer objects between different activities within your app. Although each Recognizer object, and each its Result object implements Parcelable interface, it is not so straightforward to put those objects into Intent and pass them around between your activities and services for two main reasons:

• Result object is tied to its Recognizer object, which manages lifetime of the native Result object.
• Result object often contains large data blocks, such as images, which cannot be transferred via Intent because of Android's Intent transaction data limit.

Although the first problem can be easily worked around by making a copy of the Result and transfer it independently, the second problem is much tougher to cope with. This is where, RecognizerBundle's methods saveToIntent and loadFromIntent come to help, as they ensure the safe passing of Recognizer objects bundled within RecognizerBundle between activities according to policy defined with method setIntentDataTransferMode:

• if set to STANDARD, the Recognizer objects will be passed via Intent using normal Intent transaction mechanism, which is limited by Android's Intent transaction data limit. This is same as manually putting Recognizer objects into Intent and is OK as long as you do not use Recognizer objects that produce images or other large objects in their Results.
• if set to OPTIMISED, the Recognizer objects will be passed via internal singleton object and no serialization will take place. This means that there is no limit to the size of data that is being passed. This is also the fastest transfer method, but it has a serious drawback - if Android kills your app to save memory for other apps and then later restarts it and redelivers Intent that should contain Recognizer objects, the internal singleton that should contain saved Recognizer objects will be empty and data that was being sent will be lost. You can easily provoke that condition by choosing No background processes under Limit background processes in your device's Developer options, and then switch from your app to another app and then back to your app.
• if set to PERSISTED_OPTIMISED, the Recognizer objects will be passed via internal singleton object (just like in OPTIMISED mode) and will additionaly be serialized into a file in your application's private folder. In case Android restarts your app and internal singleton is empty after re-delivery of the Intent, the data will be loaded from file and nothing will be lost. The files will be automatically cleaned up when data reading takes place. Just like OPTIMISED, this mode does not have limit to the size of data that is being passed and does not have a drawback that OPTIMISED mode has, but some users might be concerned about files to which data is being written.
  • These files will contain end-user's private data, such as image of the object that was scanned and the extracted data. Also these files may remain saved in your application's private folder until the next successful reading of data from the file.
  • If your app gets restarted multiple times, only after first restart will reading succeed and will delete the file after reading. If multiple restarts take place, you must implement onSaveInstanceState and save bundle back to file by calling its saveState method. Also, after saving state, you should ensure that you clear saved state in your onResume, as onCreate may not be called if activity is not restarted, while onSaveInstanceState may be called as soon as your activity goes to background (before onStop), even though activity may not be killed at later time.
  • If saving data to file in private storage is a concern to you, you should use either OPTIMISED mode to transfer large data and image between activities or create your own mechanism for data transfer. Note that your application's private folder is only accessible by your application and your application alone, unless the end-user's device is rooted.

List of available recognizers

This section will give a list of all Recognizer objects that are available within PhotoPay SDK, their purpose and recommendations how they should be used to get best performance and user experience.

Frame Grabber Recognizer

The FrameGrabberRecognizer is the simplest recognizer in PhotoPay SDK, as it does not perform any processing on the given image, instead it just returns that image back to its FrameCallback. Its Result never changes state from Empty.

This recognizer is best for easy capturing of camera frames with RecognizerRunnerView. Note that Image sent to onFrameAvailable are temporary and their internal buffers all valid only until the onFrameAvailable method is executing - as soon as method ends, all internal buffers of Image object are disposed. If you need to store Image object for later use, you must create a copy of it by calling clone.

Also note that FrameCallback interface extends Parcelable interface, which means that when implementing FrameCallback interface, you must also implement Parcelable interface.

This is especially important if you plan to transfer FrameGrabberRecognizer between activities - in that case, keep in mind that the instance of your object may not be the same as the instance on which onFrameAvailable method gets called - the instance that receives onFrameAvailable calls is the one that is created within activity that is performing the scan.

Success Frame Grabber Recognizer

The SuccessFrameGrabberRecognizer is a special Recognizer that wraps some other Recognizer and impersonates it while processing the image. However, when the Recognizer being impersonated changes its Result into Valid state, the SuccessFrameGrabberRecognizer captures the image and saves it into its own Result object.

Since SuccessFrameGrabberRecognizer impersonates its slave Recognizer object, it is not possible to give both concrete Recognizer object and SuccessFrameGrabberRecognizer that wraps it to same RecognizerBundle - doing so will have the same result as if you have given two instances of same Recognizer type to the RecognizerBundle - it will crash your application.

This recognizer is best for use cases when you need to capture the exact image that was being processed by some other Recognizer object at the time its Result became Valid. When that happens, SuccessFrameGrabber's Result will also become Valid and will contain described image. That image can then be retrieved with getSuccessFrame() method.

PhotoPay recognizers

Unless stated otherwise for concrete recognizer, recognizers from this list can be used in any context, but they work best with the PhotopayActivity, which has the UI best suited for scanning payment slips and payment barcodes.

SEPA Payment QR code recognizer

The SepaQrCodePaymentRecognizer is used for scanning payment information from SEPA (Single Euro Payments Area) payment QR codes. The recognizer support scanning payment QR codes that are encoded by standard defined by European Payments Council.

Country-specific PhotoPay recognizers

Austria

Austrian payslip recognizer

The AustriaSlipRecognizer is used for scanning payment information from austrian payment slips. It supports both SEPA and old version od the payment slip standard. This recognizer works only in landscape orientation, so we recommend using the PhotopayActivity with that recognizer, as it will automatically switch to landscape-only orientation when this recognizer is used.

Austrian payment QR code recognizer

The AustriaQrCodeRecognizer is used for scanning payment information from QR code usually found on SEPA payment slips in Austria.

Belgium

Belgian payslip recognizer

The 'BelgiumSlipRecognizer' is used for scanning payment information from belgian payment slips. This recognizer works only in landscape orientation, so we recommend using the PhotopayActivity with that recognizer, as it will automatically switch to landscape-only orientation when this recognizer is used.

Croatia

Croatian payslip recognizer

The CroatiaSlipRecognizer is used for scanning payment information from croatian payment slips. It supports both HUB3 and old HUB1 version od the payment slip standard.

Croatian payment PDF417 2D barcode recognizer

The CroatiaPdf417PaymentRecognizer is used for scanning payment information from PDF417 2D barcode usually found on payment slips. It supports both HUB3 and HUB1 2D barcode standards.

Croatian payment QR code recognizer

The CroatiaQrCodePaymentRecognizer is used for scanning payment information from QR codes that have content encoded in same format as specified by HUB3 PDF417 2D barcode standard.

Czechia

Czech payslip recognizer

The CzechiaSlipRecognizer is used for scanning payment information from czech payment slips.

Czech payment QR code recognizer

The CzechiaQrCodeRecognizer is used for scanning payment information from payment QR codes that are usually found on czech payment slips.

Germany

German payslip recognizer

The GermanySlipRecognizer is used for scanning payment information from german payment slips. It supports both SEPA and old version od the payment slip standard. This recognizer works only in landscape orientation, so we recommend using the PhotopayActivity with that recognizer, as it will automatically switch to landscape-only orientation when this recognizer is used.

German payment QR code recognizer

The GermanyQrCodeRecognizer is used for scanning payment information from QR code usually found on SEPA payment slips in Germany.

Hungary

Hungarian payslip recognizer

The HungarySlipRecognizer is used for scanning payment information from hungarian payment slips. It supports both yellow and white versions od the payment slip standard. This recognizer works only in landscape orientation, so we recommend using the PhotopayActivity with that recognizer, as it will automatically switch to landscape-only orientation when this recognizer is used.

Kosovo

Kosovo payslip recognizer

The KosovoSlipRecognizer is used for scanning payment information from payment slips in Kosovo. This recognizer works by scanning OCR line from the lower part of the payment slip, so we recommend using the OcrLineScanActivity with that recognizer, as it has UI best suited for scanning OCR lines.

Kosovo Code 128 recognizer

The KosovoCode128Recognizer is used for scanning payment information from Code128 1D barcodes usually found on payment slips in Kosovo.

Netherlands

Dutch payslip recognizer

The NetherlandsSlipRecognizer is used for scanning payment information from payment slips in Netherlands. This recognizer works by scanning OCR line from the lower part of the payment slip, so we recommend using the OcrLineScanActivity with that recognizer, as it has UI best suited for scanning OCR lines.

Serbia

Serbian payment PDF417 2D barcode recognizer

The SerbiaPdf417PaymentRecognizer is used for scanning payment information from PDF417 2D barcode found on some serbian invoices. The Republic of Serbia does not have a national standard for payment slips nor payment barcodes. This recognizer supports scanning PDF417 2D barcodes that are modelled after Croatian HUB3 standard.

Serbian payment QR code recognizer

The SerbiaQrCodePaymentRecognizer is used for scanning payment information from QR code found on some serbian invoices. The Republic of Serbia does not have a national standard for payment slips nor payment barcodes. This recognizer supports scanning QR codes that are modelled after Croatian HUB3 standard.

Slovakia

Slovak payslip recognizer

The SlovakiaSlipRecognizer is used for scanning payment information from payment slips in Slovakia. It supports only green payslips. You cannot use this recognizer to perform OCR of white slips. This recognizer works only in landscape orientation, so we recommend using the PhotopayActivity with that recognizer, as it will automatically switch to landscape-only orientation when this recognizer is used.

Slovak payment Code 128 recognizer

The SlovakiaCode128PaymentRecognizer is used for scanning payment information from Code128 1D barcode usually found on both white and green payment slips in Slovakia.

Slovak payment Data Matrix Code recognizer

The SlovakiaDataMatrixPaymentRecognizer is used for scanning payment information from Data Matrix 2D barcode usually found on some white payment slips in Slovakia.

Slovak payBySquare QR code recognizer

The SlovakiaQrCodePaymentRecognizer is used for scanning payment information from Slovak pyBySquare payment QR code. This recognizer support only scanning the blue (PAY bySquare) QR codes. The orange (INVOICE bySquare) QR codes are not supported by this recognizer.

Slovenia

Slovenian payslip recognizer

The SloveniaSlipRecognizer is used for scanning payment information from slovenian UPN payment slips. It performs the OCR of the left part of the slip (the talon) - it is not able to perform scanning of right part of the slip nor the OCR line.

Slovenian payment QR code recognizer

The SloveniaQrCodePaymentRecognizer is used for scanning payment information from payment QR codes usually found on UPN payment slips in Slovenia.

Switzerland

Swiss payslip recognizer

The SwitzerlandSlipRecognizer is used for scanning payment information from payment slips in Switzerland. It supports scanning only OCR lines in orange slips. It does not support red payment slips. This recognizer works by scanning OCR line from the lower part of the payment slip, so we recommend using the OcrLineScanActivity with that recognizer, as it has UI best suited for scanning OCR lines.

Swiss payment QR code recognizer

The SwitzerlandQrCodePaymentRecognizer is used for scanning payment information from payment QR codes used in Switzerland.

United Kingdom

UK payslip recognizer

The UnitedKingdomSlipRecognizer is used for scanning payment information from payment slips in United Kingdom. It supports scanning both OCR and MICR lines found in UK giro credit slips. This recognizer works by scanning OCR line from the lower part of the payment slip, so we recommend using the OcrLineScanActivity with that recognizer, as it has UI best suited for scanning OCR lines.

UK payment QR code recognizer

The UnitedKingdomQrCodePaymentRecognizer is used for scanning payment information from payment QR codes in United Kingdom. Please contact us for more information about supported QR code standards.

BlinkInput recognizer

The BlinkInputRecognizer is generic OCR recognizer used for scanning segments which enables specifying Processors that will be used for scanning. Most commonly used Processor within this recognizer is ParserGroupProcessor that activates all Parsers in the group to extract data of interest from the OCR result.

This recognizer can be used in any context. It is used internally in the implementation of the provided FieldByFieldOverlayController.

Processors are explained in The Processor concept section and you can find more about Parsers in The Parser concept section.

Field by field scanning feature

Field by field scanning feature is designed for scanning small text fields which are called scan elements. Elements are scanned in the predefined order. For each scan element, specific Parser that will extract structured data of interest from the OCR result is defined. Focusing on the small text fields which are scanned one by one enables implementing support for the free-form documents because field detection is not required. The user is responsible for positioning the field of interest inside the scanning window and the scanning process guides him. When implementing support for the custom document, only fields of interest has to be defined.

Field by field scan can be performed by using provided FieldByFieldScanActivity and FieldByFieldOverlayController.

For preparing the scan configuration, FieldByFieldBundle is used. It holds the array of FieldByFieldElements passed to its constructor and it is responsible for transferring them from one Activity to another, just like the RecognizerBundle transfers Recognizers.

FieldByFieldElement holds a combination of Parser used for data extraction, its title and message which are shown in the UI during the scan. For all available configuration options please see the javadoc.

When FieldByFieldBundle is prepared, it should be used for creating the FieldByFieldUISettings which accepts FieldByFieldBundle as a constructor argument and can be used to additionally tweak the scanning process and UI. For the list of all available configuration options, please see javadoc.

For starting the FieldByFieldScanActivity, the ActivityRunner.startActivityForResult should be called with the prepared FieldByFieldUISettings.

When the scanning is done and control is returned to the calling activity, in onActivityResult method FieldByFieldBundle.loadFromIntent should be called. FieldByFieldBundle will load the scanning results to the Parser instances held by its elements.

Performing your first field by field scan

1. Before starting a recognition process, you need to obtain a license from Microblink dashboard. After registering, you will be able to generate a trial license for your app. License is bound to package name of your app, so please make sure you enter the correct package name when asked.

   After creating a license, you will have the option to download the license as a file that you must place within your application's assets folder. You must ensure that license key is set before instantiating any other classes from the SDK, otherwise you will get an exception at runtime. Therefore, we recommend that you extend Android Application class and set the license in its onCreate callback in the following way:

   public class MyApplication extends Application {
       @Override
       public void onCreate() {
           MicroblinkSDK.setLicenseFile("path/to/license/file/within/assets/dir", this);
       }
   }

2. In your main activity create parser objects that will be used during recognition, configure them if needed, define scan elements and store them in FieldByFieldBundle object. For example, to scan three fields: amount, date and raw text, you can configure your recognizer object in the following way:

    public class MyActivity extends Activity {
        // parsers are member variables because it will be used for obtaining results
        private AmountParser mAmountParser;
        private DateParser mDateParser;
        private RawParser mRawParser;
   
        /** Reference to bundle is kept, it is used later for loading results from intent */
        private FieldByFieldBundle mFieldByFieldBundle;
        
        @Override
        protected void onCreate(Bundle bundle) {
            super.onCreate(bundle);
            
            // setup views, as you would normally do in onCreate callback
            
            mAmountParser = new AmountParser();
            mDateParser = new DateParser();
            mRawParser = new RawParser();
            
            // prepare scan elements and put them in FieldByFieldBundle
            // we need to scan 3 items, so we will create bundle with 3 elements
            mFieldByFieldBundle = new FieldByFieldBundle(
                // each scan element contains two string resource IDs: string shown in title bar
                // and string shown in text field above scan box. Besides that, it contains parser
                // that will extract data from the OCR result.
                new FieldByFieldElement(R.string.amount_title, R.string.amount_msg, mAmountParser),
                new FieldByFieldElement(R.string.date_title, R.string.date_msg, mDateParser),
                new FieldByFieldElement(R.string.raw_title, R.string.raw_msg, mRawParser)
            );
        }
    }

3. You can start recognition process by starting FieldByFieldScanActivity. You need to do that by creating FieldByFieldUISettings and calling ActivityRunner.startActivityForResult method:

    // method within MyActivity from previous step
    public void startFieldByFieldScanning() {
        // we use FieldByFieldUISettings - settings for FieldByFieldScanActivity
        FieldByFieldUISettings scanActivitySettings = new FieldByFieldUISettings(mFieldByFieldBundle);
        
        // tweak settings as you wish
        
        // Start activity
        ActivityRunner.startActivityForResult(this, MY_REQUEST_CODE, scanActivitySettings);
    }

4. After FieldByFieldScanActivity finishes the scan, it will return to the calling activity or fragment and will call its method onActivityResult. You can obtain the scanning results in that method.

   @Override
   protected void onActivityResult(int requestCode, int resultCode, Intent data) {
       super.onActivityResult(requestCode, resultCode, data);
       
       if (requestCode == MY_REQUEST_CODE) {
           if (resultCode == FieldByFieldScanActivity.RESULT_OK && data != null) {
               // load the data into all parsers bundled within your FieldByFieldBundle
               mFieldByFieldBundle.loadFromIntent(data);
               
               // now every parser object that was bundled within FieldByFieldBundle
               // has been updated with results obtained during scanning session
               
               // you can get the results by invoking getResult on each parser, and then
               // invoke specific getter for each concrete parser result type
               String amount = mAmountParser.getResult().getAmount();
               String date = mDateParser.getResult().getDate().toString();
               String rawText = mRawParser.getResult().getRawText();
   
               if (!amount.isEmpty()) {
                   // amount has been successfully parsed, you can use it however you wish
               }
               if (!date.isEmpty()) {
                   // date has been successfully parsed, you can use it however you wish
               }
               if (!rawText.isEmpty()) {
                   // raw text has been successfully returned, you can use it however you wish
               }
           }
       }
   }

Processor and Parser

The Processors and Parsers are standard processing units within PhotoPay SDK used for data extraction from the input images. Unlike the Recognizer, Processor and Parser are not stand-alone processing units. Processor is always used within Recognizer and Parser is used within appropriate Processor to extract data from the OCR result.

The Processor concept

Processor is a processing unit used within some Recognizer which supports processors. It processes the input image prepared by the enclosing Recognizer in the way that is characteristic to the implementation of the concrete Processor.

For example, BlinkInputRecognizer encloses a collection of processors which are run on the input image to extract data. To perform the OCR of the input image, ParserGroupProcessor is used.

Processor architecture is similar to Recognizer architecture described in The Recognizer concept section. Each instance also has associated inner Result object whose lifetime is bound to the lifetime of its parent Processor object and it is updated while Processor works. If you need your Result object to outlive its parent Processor object, you must make a copy of it by calling its method clone().

It also has its internal state and while it is in the working state during recognition process, it is not allowed to tweak Processor object's properties.

To support common use cases, there are several different Processor implementations available. They are listed in the next section.

List of available processors

This section will give a list of Processor types that are available within PhotoPay SDK and their purpose.

Parser Group Processor

The ParserGroupProcessor is the type of the processor that performs the OCR (Optical Character Recognition) on the input image and lets all the parsers within the group to extract data from the OCR result. The concept of Parser is described in the next section.

Before performing the OCR, the best possible OCR engine options are calculated by combining engine options needed by each Parser from the group. For example, if one parser expects and produces result from uppercase characters and other parser extracts data from digits, both uppercase characters and digits must be added to the list of allowed characters that can appear in the OCR result. This is a simplified explanation because OCR engine options contain many parameters which are combined by the ParserGroupProcessor.

Because of that, if multiple parsers and multiple parser group processors are used during the scan, it is very important to group parsers carefully.

Let's see this on an example: assume that we have two parsers at our disposal: AmountParser and IbanParser. AmountParser knows how to extract amount's from OCR result and requires from OCR only to recognize digits, periods and commas and ignore letters. On the other hand, IbanParser knows how to extract IBAN from OCR result and requires from OCR to recognize letters and digits.

If we put both AmountParser and IbanParser into the same ParserGroupProcessor, the merged OCR engine settings will require recognition of all letters, all digits, both period and comma. Such OCR result will contain all characters for IbanParser to properly parse IBAN, but might confuse AmountParser if OCR misclassifies some characters into digits.

If we put AmountParser in one ParserGroupProcessor and IbanParser in another ParserGroupProcessor, OCR will be performed for each parser group independently, thus preventing the AmountParser confusion, but two OCR passes of the image will be performed, which can have a performance impact.

ParserGroupProcessor is most commonly used Processor. It is used whenever the OCR is needed. After the OCR is performed and all parsers are run, parsed results can be obtained through parser objects that are enclosed in the group. ParserGroupProcessor instance also has associated inner ParserGroupProcessor.Result whose state is updated during processing and its method getOcrResult() can be used to obtain the raw OCRResult that was used for parsing data.

Take note that OCRResult is available only if it is allowed by the PhotoPay SDK license key. OCRResult structure contains information about all recognized characters and their positions on the image. To prevent someone to abuse that, obtaining of the OCRResult structure is allowed only by the premium license keys.

The Parser concept

Parser is a class of objects that are used to extract structured data from the raw OCR result. It must be used within ParserGroupProcessor which is responsible for performing the OCR, so Parser is not stand-alone processing unit.

Like Recognizer and all other processing units, each Parser instance has associated inner Result object whose lifetime is bound to the lifetime of its parent Parser object and it is updated while Parser works. When parsing is done Result can be used for obtaining extracted data. If you need your Result object to outlive its parent Parser object, you must make a copy of it by calling its method clone().

It also has its internal state and while it is in the working state during recognition process, it is not allowed to tweak Parser object's properties.

There are a lot of different Parsers for extracting most common fields which appear on various documents. Also, most of them can be adjusted for specific use cases.

List of available parsers

Amount Parser

AmountParser is used for extracting amounts from the OCR result. For available configuration options and result getters please check javadoc.

Date Parser

DateParser is used for extracting dates in various formats from the OCR result. For available configuration options and result getters please check javadoc.

IBAN Parser

IbanParser is used for extracting IBAN (International Bank Account Number) from the OCR result. For available configuration options and result getters please check javadoc.

Raw Parser

RawParser is used for obtaining string version of raw OCR result, without performing any smart parsing operations. For available result getters please check javadoc.

List of available parsers for payment information scanning by countries

Australia

Australian Business Number parser

The AustraliaAbnParser is used for scanning Australian Business Number.

Australian Account Number parser

The AustraliaAccountParser is used for scanning Australian Account Number.

Australian Bank State Branch parser

The AustraliaBsbParser is used for scanning Australian Bank State Branch.

Australian Biller Code parser

The AustraliaBillerParser is used for scanning Australian Biller Code.

Australian payment reference number parser

The AustraliaReferenceParser is used for scanning Australian payment reference number.

Austria

Austrian payment reference number parser

The AustriaReferenceParser is used for scanning austrian 12-digit payment reference number (Kundendaten).

Bosnia and Herzegovina

Account number from Bosnia and Herzegovina parser

The BosniaAndHerzegovinaAccountParser is used for scanning account number from Bosnia and Herzegovina.

Payment reference number from Bosnia and Herzegovina parser

The BosniaAndHerzegovinaReferenceParser is used for scanning payment reference number from Bosnia and Herzegovina.

Croatia

Croatian amount parser

The CroatiaAmountParser is used for scanning amounts from payment slips and invoices issued in Croatia.

Croatian payment reference number parser

The CroatiaReferenceParser is used for scanning Croatian payment reference numbers.

Czechia

Czech account number parser

The CzechiaAccountParser is used for scanning Czech account numbers.

Czech Variabilni Symbol parser

The CzechiaVariabilniSymbolParser is used for scanning Czech variabilni symbol number.

Germany

German payment reference number parser

The GermanyReferenceParser is used for scanning both German 13-digit payment reference numbers with ISO 7064 check digits as well as RF payment reference numbers.

Hungary

Hungarian account number parser

The HungaryAccountParser is used for scanning hungarian account number.

Hungarian payer ID number parser

The HungaryPayerIdParser is used for scanning hungarian payer ID number.

Macedonia

Macedonian account number parser

The MacedoniaAccountParser is used for scanning macedonian account numbers.

Macedonian payment reference number parser

The MacedoniaReferenceParser is used for scanning Macedonian payment reference numbers.

Montenegro

Montenegro account number parser

The MontenegroAccountParser is used for scanning account numbers in Montenegro.

Montenegro payment reference number parser

The MontenegroReferenceParser is used for scanning payment reference numbers in Montenegro.

Serbia

Serbian account number parser

The SerbiaAccountParser is used for scanning serbian account numbers.

Serbian payment reference number parser

The SerbiaReferenceParser is used for scanning serbian payment reference numbers.

Slovenia

Slovenian payment reference number parser

The SloveniaReferenceParser is used for scanning slovenian payment reference numbers.

Sweden

Swedish amount parser

The SwedenAmountParser is used for scanning amount on swedish payment slip.

Swedish Giro number parser

The SwedenGiroNumberParser is used for scanning giro number on swedish payment slip.

Swedish payment reference number parser

The SwedenReferenceParser is used for scanning payment reference number on swedish payment slip.

Swedish slip code number parser

The SwedenSlipCodeParser is used for scanning slip code number on swedish payment slip.

PDF417 recognizer

The Pdf417Recognizer is recognizer specialised for scanning PDF417 2D barcodes. This recognizer can recognize only PDF417 2D barcodes - for recognition of other barcodes, please refer to BarcodeRecognizer.

This recognizer can be used in any context, but it works best with the BarcodeScanActivity, which has UI best suited for barcode scanning.

Barcode recognizer

The BarcodeRecognizer is recognizer specialised for scanning various types of barcodes. This recognizer should be your first choice when scanning barcodes as it supports lots of barcode symbologies, including the PDF417 2D barcodes, thus making PDF417 recognizer possibly redundant, which was kept only for its simplicity.

As you can see from javadoc, you can enable multiple barcode symbologies within this recognizer, however keep in mind that enabling more barcode symbologies affects scanning performance - the more barcode symbologies are enabled, the slower the overall recognition performance. Also, keep in mind that some simple barcode symbologies that lack proper redundancy, such as Code 39, can be recognized within more complex barcodes, especially 2D barcodes, like PDF417.

This recognizer can be used in any context, but it works best with the BarcodeScanActivity, which has UI best suited for barcode scanning.

Embedding PhotoPay inside another SDK

You need to ensure that the final app gets all resources required by PhotoPay. At the time of writing this documentation, Android does not have support for combining multiple AAR libraries into single fat AAR. The problem is that resource merging is done while building application, not while building AAR, so application must be aware of all its dependencies. There is no official Android way of "hiding" third party AAR within your AAR.

This problem is usually solved with transitive Maven dependencies, i.e. when publishing your AAR to Maven you specify dependencies of your AAR so they are automatically referenced by app using your AAR. Besides this, there are also several other approaches you can try:

• you can ask your clients to reference PhotoPay in their app when integrating your SDK
• since the problem lies in resource merging part you can try avoiding this step by ensuring your library will not use any component from PhotoPay that uses resources (i.e. built-in activities, fragments and views, except RecognizerRunnerView). You can perform custom UI integration while taking care that all resources (strings, layouts, images, ...) used are solely from your AAR, not from PhotoPay. Then, in your AAR you should not reference LibPhotoPay.aar as gradle dependency, instead you should unzip it and copy its assets to your AAR’s assets folder, its classes.jar to your AAR’s lib folder (which should be referenced by gradle as jar dependency) and contents of its jni folder to your AAR’s src/main/jniLibs folder.
• Another approach is to use 3rd party unofficial gradle script that aim to combine multiple AARs into single fat AAR. Use this script at your own risk and report issues to its developers - we do not offer support for using that script.
• There is also a 3rd party unofficial gradle plugin which aims to do the same, but is more up to date with latest updates to Android gradle plugin. Use this plugin at your own risk and report all issues with using to its developers - we do not offer support for using that plugin.

Processor architecture considerations

PhotoPay is distributed with ARMv7 and ARM64 native library binaries.

ARMv7 architecture gives the ability to take advantage of hardware accelerated floating point operations and SIMD processing with NEON. This gives PhotoPay a huge performance boost on devices that have ARMv7 processors. Most new devices (all since 2012.) have ARMv7 processor so it makes little sense not to take advantage of performance boosts that those processors can give. Also note that some devices with ARMv7 processors do not support NEON and VFPv4 instruction sets, most popular being those based on NVIDIA Tegra 2, ARM Cortex A9 and older. Since these devices are old by today's standard, PhotoPay does not support them. For the same reason, PhotoPay does not support devices with ARMv5 (armeabi) architecture.

ARM64 is the new processor architecture that most new devices use. ARM64 processors are very powerful and also have the possibility to take advantage of new NEON64 SIMD instruction set to quickly process multiple pixels with a single instruction.

There are some issues to be considered:

• ARMv7 build of the native library cannot be run on devices that do not have ARMv7 compatible processor
• ARMv7 processors do not understand x86 instruction set
• ARM64 processors understand ARMv7 instruction set, but ARMv7 processors do not understand ARM64 instructions.
  • NOTE: as of the year 2018, some android devices that ship with ARM64 processors do not have full compatibility with ARMv7. This is mostly due to incorrect configuration of Android's 32-bit subsystem by the vendor, however Google decided that as of August 2019 all apps on PlayStore that contain native code need to have native support for 64-bit processors (this includes ARM64 and x86_64) - this is in anticipation of future Android devices that will support 64-bit code only, i.e. that will have ARM64 processors that do not understand ARMv7 instruction set.
• if ARM64 processor executes ARMv7 code, it does not take advantage of modern NEON64 SIMD operations and does not take advantage of 64-bit registers it has - it runs in emulation mode

LibPhotoPay.aar archive contains ARMv7 and ARM64 builds of the native library. By default, when you integrate PhotoPay into your app, your app will contain native builds for all these processor architectures. Thus, PhotoPay will work on ARMv7 and ARM64 devices and will use ARMv7 features on ARMv7 devices and ARM64 features on ARM64 devices. However, the size of your application will be rather large.

Reducing the final size of your app

We recommend that you distribute your app using App Bundle. This will defer apk generation to Google Play, allowing it to generate minimal APK for each specific device that downloads your app, including only required processor architecture support.

Using APK splits

If you are unable to use App Bundle, you can create multiple flavors of your app - one flavor for each architecture. With gradle and Android studio this is very easy - just add the following code to build.gradle file of your app:

android {
  ...
  splits {
    abi {
      enable true
      reset()
      include 'armeabi-v7a', 'arm64-v8a'
      universalApk true
    }
  }
}

With that build instructions, gradle will build two different APK files for your app. Each APK will contain only native library for one processor architecture and one APK will contain all architectures. In order for Google Play to accept multiple APKs of the same app, you need to ensure that each APK has different version code. This can easily be done by defining a version code prefix that is dependent on architecture and adding real version code number to it in following gradle script:

// map for the version code
def abiVersionCodes = ['armeabi-v7a':1, 'arm64-v8a':2]

import com.android.build.OutputFile

android.applicationVariants.all { variant ->
    // assign different version code for each output
    variant.outputs.each { output ->
        def filter = output.getFilter(OutputFile.ABI)
        if(filter != null) {
            output.versionCodeOverride = abiVersionCodes.get(output.getFilter(OutputFile.ABI)) * 1000000 + android.defaultConfig.versionCode
        }
    }
}

For more information about creating APK splits with gradle, check this article from Google.

After generating multiple APK's, you need to upload them to Google Play. For tutorial and rules about uploading multiple APK's to Google Play, please read the official Google article about multiple APKs.

Removing processor architecture support

If you won't be distributing your app via Google Play or for some other reasons want to have single APK of smaller size, you can completely remove support for certain CPU architecture from your APK. This is not recommended due to consequences.

To keep only some CPU architectures, for example armeabi-v7a and arm64-v8a, add the following statement to your android block inside build.gradle:

android {
    ...
    ndk {
        // Tells Gradle to package the following ABIs into your application
        abiFilters 'armeabi-v7a', 'arm64-v8a'
    }
}

This will remove other architecture builds for all native libraries used by the application.

To remove support for a certain CPU architecture only for PhotoPay, add the following statement to your android block inside build.gradle:

android {
    ...
    packagingOptions {
        exclude 'lib/<ABI>/libBlinkPhotoPay.so'
    }
}

where <ABI> represents the CPU architecture you want to remove:

• to remove ARMv7 support, use exclude 'lib/armeabi-v7a/libBlinkPhotoPay.so'
• to remove ARM64 support, use exclude 'lib/arm64-v8a/libBlinkPhotoPay.so'
  • NOTE: this is not recommended. See this notice.

You can also remove multiple processor architectures by specifying exclude directive multiple times. Just bear in mind that removing processor architecture will have side effects on performance and stability of your app. Please read this for more information.

Consequences of removing processor architecture

• Google decided that as of August 2019 all apps on Google Play that contain native code need to have native support for 64-bit processors (this includes ARM64 and x86_64). This means that you cannot upload application to Google Play Console that supports only 32-bit ABI and does not support corresponding 64-bit ABI.

• By removing ARMv7 support, PhotoPay will not work on devices that have ARMv7 processors.

• By removing ARM64 support, PhotoPay will not use ARM64 features on ARM64 device

  • also, some future devices may ship with ARM64 processors that will not support ARMv7 instruction set. Please see this note for more information.

Combining PhotoPay with other native libraries

If you are combining PhotoPay library with other libraries that contain native code into your application, make sure you match the architectures of all native libraries. For example, if third party library has got only ARMv7 version, you must use exactly ARMv7 version of PhotoPay with that library, but not ARM64. Using this architectures will crash your app at initialization step because JVM will try to load all its native dependencies in same preferred architecture and will fail with UnsatisfiedLinkError.

Resolving conflict on libc++_shared.so

PhotoPay contains native code that depends on the C++ runtime. This runtime is provided by the libc++_shared.so, which needs to be available in your app that is using PhotoPay. However, the same file is also used by various other libraries that contain native components. If you happen to integrate both such library together with PhotoPay in your app, your build will fail with an error similar to this one:

* What went wrong:
Execution failed for task ':app:mergeDebugNativeLibs'.
> A failure occurred while executing com.android.build.gradle.internal.tasks.MergeJavaResWorkAction
   > 2 files found with path 'lib/arm64-v8a/libc++_shared.so' from inputs:
      - <path>/.gradle/caches/transforms-3/3d428f9141586beb8805ce57f97bedda/transformed/jetified-opencv-4.5.3.0/jni/arm64-v8a/libc++_shared.so
      - <path>/.gradle/caches/transforms-3/609476a082a81bd7af00fd16a991ee43/transformed/jetified-photopay-9.1.2/jni/arm64-v8a/libc++_shared.so
     If you are using jniLibs and CMake IMPORTED targets, see
     https://developer.android.com/r/tools/jniLibs-vs-imported-targets

The error states that multiple different dependencies provide the same file lib/arm64/libc++_shared.so (in this case, OpenCV and PhotoPay).

You can resolve this issue by making sure that the dependency that uses newer version of libc++_shared.so is listed first in your dependency list, and then, simply add the following to your build.gradle:

android {
    packaging {
        jniLibs {
            pickFirsts.add("lib/armeabi-v7a/libc++_shared.so")
            pickFirsts.add("lib/arm64-v8a/libc++_shared.so")
        }
    }
}

IMPORTANT NOTE

The code above will always select the first libc++_shared.so from your dependency list, so make sure that the dependency that uses the latest version of libc++_shared.so is listed first. This is because libc++_shared.so is backward-compatible, but not forward-compatible. This means that, e.g. libBlinkPhotoPay.so built against libc++_shared.so from NDK r24 will work without problems when you package it together with libc++_shared.so from NDK r26, but will crash when you package it together with libc++_shared.so from NDK r21. This is true for all your native dependencies.

Troubleshooting

Integration difficulties

In case of problems with SDK integration, first make sure that you have followed integration instructions. If you're still having problems, please contact us at help.microblink.com.

Licensing issues

If you are getting "invalid license key" error or having other license-related problems (e.g. some feature is not enabled that should be or there is a watermark on top of camera), first check the ADB logcat. All license-related problems are logged to error log so it is easy to determine what went wrong.

When you have to determine what is the license-relate problem or you simply do not understand the log, you should contact us help.microblink.com. When contacting us, please make sure you provide following information:

• exact package name of your app (from your AndroidManifest.xml and/or your build.gradle file)
• license that is causing problems
• please stress out that you are reporting problem related to Android version of PhotoPay SDK
• if unsure about the problem, you should also provide excerpt from ADB logcat containing license error

Keep in mind: Versions 7.11.0 and above require an internet connection to work under our new License Management Program.

We’re only asking you to do this so we can validate your trial license key. Data extraction still happens offline, on the device itself. Once the validation is complete, you can continue using the SDK in offline mode (or over a private network) until the next check.

Other problems

If you are having problems with scanning certain items, undesired behaviour on specific device(s), crashes inside PhotoPay or anything unmentioned, please do as follows:

• enable logging to get the ability to see what is library doing. To enable logging, put this line in your application:

  com.microblink.photopay.util.Log.setLogLevel(com.microblink.photopay.util.Log.LogLevel.LOG_VERBOSE);

  After this line, library will display as much information about its work as possible. Please save the entire log of scanning session to a file that you will send to us. It is important to send the entire log, not just the part where crash occurred, because crashes are sometimes caused by unexpected behaviour in the early stage of the library initialization.

• Contact us at help.microblink.com describing your problem and provide following information:

  • log file obtained in previous step
  • high resolution scan/photo of the item that you are trying to scan
  • information about device that you are using - we need exact model name of the device. You can obtain that information with any app like this one
  • please stress out that you are reporting problem related to Android version of PhotoPay SDK

FAQ and known issues

After switching from trial to production license I get InvalidLicenseKeyException when I construct specific Recognizer object

Each license key contains information about which features are allowed to use and which are not. This exception indicates that your production license does not allow using of specific Recognizer object. You should contact support to check if provided license is OK and that it really contains all features that you have purchased.

I get InvalidLicenseKeyException with trial license key

Whenever you construct any Recognizer object or any other object that derives from Entity, a check whether license allows using that object will be performed. If license is not set prior constructing that object, you will get InvalidLicenseKeyException. We recommend setting license as early as possible in your app, ideally in onCreate callback of your Application singleton.

When my app starts, I get exception telling me that some resource/class cannot be found or I get ClassNotFoundException

This usually happens when you perform integration into Eclipse project and you forget to add resources or native libraries into the project. You must alway take care that same versions of both resources, assets, java library and native libraries are used in combination. Combining different versions of resources, assets, java and native libraries will trigger crash in SDK. This problem can also occur when you have performed improper integration of PhotoPay SDK into your SDK. Please read how to embed PhotoPay inside another SDK.

When my app starts, I get UnsatisfiedLinkError

This error happens when JVM fails to load some native method from native library If performing integration into Android studio and this error happens, make sure that you have correctly combined PhotoPay SDK with third party SDKs that contain native code, especially if you need resolving conflict over libc++_shared.so. If this error also happens in our integration sample apps, then it may indicate a bug in the SDK that is manifested on specific device. Please report that to our support team.

During build, I get conflict about duplicate libc++_shared.so

Please consult the section about resolving libc++_shared.so conflict.

I've added my callback to MetadataCallbacks object, but it is not being called

Make sure that after adding your callback to MetadataCallbacks you have applied changes to RecognizerRunnerView or RecognizerRunner as described in this section.

I've removed my callback to MetadataCallbacks object, and now app is crashing with NullPointerException

Make sure that after removing your callback from MetadataCallbacks you have applied changes to RecognizerRunnerView or RecognizerRunner as described in this section.

In my onScanningDone callback I have the result inside my Recognizer, but when scanning activity finishes, the result is gone

This usually happens when using RecognizerRunnerView and forgetting to pause the RecognizerRunnerView in your onScanningDone callback. Then, as soon as onScanningDone happens, the result is mutated or reset by additional processing that Recognizer performs in the time between end of your onScanningDone callback and actual finishing of the scanning activity. For more information about statefulness of the Recognizer objects, check this section.

I am using built-in activity to perform scanning and after scanning finishes, my app crashes with IllegalStateException stating Data cannot be saved to intent because its size exceeds intent limit.

This usually happens when you use Recognizer that produces image or similar large object inside its Result and that object exceeds the Android intent transaction limit. You should enable different intent data transfer mode. For more information about this, check this section. Also, instead of using built-in activity, you can use RecognizerRunnerFragment with built-in scanning overlay.

After scanning finishes, my app freezes

This usually happens when you attempt to transfer standalone Result that contains images or similar large objects via Intent and the size of the object exceeds Android intent transaction limit. Depending on the device, you will get either TransactionTooLargeException, a simple message BINDER TRANSACTION FAILED in log and your app will freeze or your app will get into restart loop. We recommend that you use RecognizerBundle and its API for sending Recognizer objects via Intent in a more safe manner (check this section for more information). However, if you really need to transfer standalone Result object (e.g. Result object obtained by cloning Result object owned by specific Recognizer object), you need to do that using global variables or singletons within your application. Sending large objects via Intent is not supported by Android.

Scanning with a camera works better than a recognition of images by using the Direct API

When automatic scanning of camera frames with our camera management is used (provided camera overlays or direct usage of RecognizerRunnerView), we use a stream of video frames and send multiple images to the recognition to boost reading accuracy. Also, we perform frame quality analysis and combine scanning results from multiple camera frames. On the other hand, when you are using the Direct API with a single image per document side, we cannot combine multiple images. We do our best to extract as much information as possible from that image. In some cases, when the quality of the input image is not good enough, for example, when the image is blurred or when glare is present, we are not able to successfully read the document.

I am getting a ‘Network required’ error when I'm on a private network

Online trial licenses require a public network access for validation purposes. See Licensing issues.

onOcrResult() method in my OcrCallback is never invoked and all Result objects always return null in their OCR result getters

In order to be able to obtain raw OCR result, which contains locations of each character, its value and its alternatives, you need to have a license that allows that. By default, licenses do not allow exposing raw OCR results in public API. If you really need that, please contact us and explain your use case.

Additional info

PhotoPay SDK size

You can find PhotoPay SDK size report for all supported ABIs here.

API reference

Complete API reference can be found in Javadoc.

Contact

For any other questions, feel free to contact us at help.microblink.com.