flutter-ml
TextRecognizer flat out fails reading embossed text.
Embossed text, such as a credit card number completely fails at being read. Printed number text works fine.
What options are available to improve the accuracy of reading embossed numbers? It's half way there, but my app almost entirely relies on the ability to read this type of font.
@codyfilatov: could you provide a sample image that is failing? I would like to take a look.
I am at 0 risk sharing this image, it's just the VISA and bank INI numbers. Nothing is PII.
Please see this image for example, it has extreme trouble identifying the card number text. Particularly the first "4" in the number. I have attempted a few things that did help slightly. I got VERY SLIGHTLY better results processing this same image as greyscale.
Which plugin are you using in your yaml file?
google_mlkit_text_recognition: 0.3.0 Using TextRecognizer getting the results with nothing fancy. ImageFile is a max resolution image taken from the flutter camera plugin library. I have absolutely 0 problems with printed card text, only embossed cards cause this issue. I can understand it is more challenging to identify them, but if there are suggestions to improve the accuracy it would help significantly.
TextRecognizer textRecognizer = TextRecognizer(script: TextRecognitionScript.latin); RecognizedText frontText = await textRecognizer.processImage(InputImage.fromFile(imageFile));
Both platforms but primary testing on android
On Sun, Jul 24, 2022, 10:44 PM Francisco Bernal @.***> wrote:
ios or android?
— Reply to this email directly, view it on GitHub https://github.com/bharat-biradar/Google-Ml-Kit-plugin/issues/301#issuecomment-1193606524, or unsubscribe https://github.com/notifications/unsubscribe-auth/AU6CENKUMIJ52EK43XFTJ3DVVYSTXANCNFSM54PIE2GQ . You are receiving this because you were mentioned.Message ID: @.***>
Thanks @c-louis, definitely it is related, if they reported the same issue in the native sample app and since this plugin is a wrapper for the native APIs using FlutterMethodChannel, then we will experience the same issue. @codyfilatov: We are going to keep this issue open, but there is nothing we can do until they fix the native libraries. Keep an eye on https://github.com/googlesamples/mlkit/issues/338.