Does the Azure FormRecognizer by default does any preprocessing to the file to improve the accuracy of recognition by the OCR Engine. Just out of curiosity I performed preprocessing(like- binarisation cv2.adaptiveThreshold, Denoised the image using cv2.fastNlMeansDenoising and cv2.medianBlur, performed Thinning as the image contains handwritten text and finally performed skew correction on the images before posting a request to azure. I sent two request one with an original image and the preprocessed image to my suprise the original image results were way-better compared to the preprocessed image. What could be the reason for this..does Azure formrecognizer do any preprocessing ? Any ideas on where did I go wrong?
Asked
Active
Viewed 301 times
1 Answers
0
Form Recognizer does all the processing, no need to pre-process images before sending them to Form Recognizer you can send the image as is and Form Recognizer will extract the data from it.
Neta
- 690
- 4
- 5