Firebase ML Kit - Character recognition support for various alphabets?

Question

Does the character recognition API contained in the Firebase ML Kit have support for different alphabets? I've tested with the English alphabet, French alphabet, and the cyrillic alphabet as well, and the characters returned as a result of executing the following code on Android:

final FirebaseVisionImage image = FirebaseVisionImage.fromBitmap(bitmap);
mTextDetector.detectInImage(image)
        .addOnSuccessListener(new OnSuccessListener<FirebaseVisionText>() {
            @Override
            public void onSuccess(FirebaseVisionText firebaseVisionText) {
                final List<FirebaseVisionText.Block> blocks = firebaseVisionText.getBlocks();
                // processing of blocks...
            }
        })
        .addOnFailureListener(new OnFailureListener() {
            @Override
            public void onFailure(@NonNull Exception e) {
               // handling of character recognition failure...
            }
        });

are always latin characters from the English alphabet, i.e. any recognized French, or cyrillic chars are replaced with latin chars which seem to match them best.

Is it possible to achieve correct recognition of such characters?

Thank you.

score 0 · Answer 1 · answered Jul 17 '18 at 22:11

Are you using the on-device or the cloud API?

The on-device model recognises only Latin characters. I am not sure about French, but I tried it with Hungarian text (Latin characters, but lots of weird accented ones), and it worked quite well (as long as the image quality was good). Also I used it in iOS, but I guess it shouldn't make a difference, the model should be the same.

https://firebase.google.com/docs/ml-kit/recognize-text

Firebase ML Kit - Character recognition support for various alphabets?

1 Answers1