Hi,
I’ve tried using OCR.space on Steam keys and it kinda failed. At first with original picture size It gave me no output; OCR skipped steam key. I’ve tried resizing picture to 50% of original size but the output was not correct. It mixed numbers and letters…
EDIT: Resized image (50% of original size) did work on google vision API but I don’t want to use that one.
Anyone got idea on how to fix that ? Here’s the original image (100% size)
Thanks in advance !
Reducing the size by 50% and having just the dotted matrix as input improves the OCR result, but we still get 6 characters wrong:
Using Greek as input OCR language reduces the number of errors to four:
OXHJH-ZLIPH-H255D
For alphanumeric OCR it can be always helpful to experiment with different OCR languages.
Indeed, Google Cloud Vision OCR is better here, but also not perfect (two letters wrong):
Some pre-processing can help: How to Improve Dot Matrix OCR Performance => If you try them, some feedback on how they worked would be appreciated 
Thanks a lot for a reply.
I have tried following that article and it did help a bit.
I’ve processed Images both in Photoshop and Python. Output was ofcourse not the same.
Both pictures got recognized at Google Vision API, but OCR.Space did fail a bit.
Also I’ve tried running tesseract OCR on images and output from Photoshop achived around 70% accuracy. I think I’ll have to get more train data for tesseract or to improve processing of image for ocr.space api which is my ideal solution.
Anyhow if you got more ideas hit me up 
Thanks !
