it doesn’t quite work that way. when you convert an image to base64 - it’s like taking each pixel’s RGB elements plus transparency and converting it to the bit notation that base64 requires. so two people handwriting a number 1, or even something as simple as a dot (period) would consist of several dozen bits and they couldn’t possibly have the same bit pattern (and that’s just a dot!). so their OCR’s won’t match and neither ot them will match the OCR of a “perfect/standard” dot somewhere for comparison. base64 conversion will never achieve that.
these two dots look the same …
…but wait till you see how well (?) the base64 translations compare to each other!
what you need is something bordering on machine learning and artificial intelligence - the reason some computers can recognize when you write the number , say 2, is because the distribution of the little pixels comprising the number is a close match of a pattern averaged out of several hundreds of thousands of people writing the same number 2. (and that doesn’t even try to match the speed and direction of the dots). matching handwritten input isn’t as simple as comparing base64 - i’m sorry to say - i don’t mean to disappoint you. but keep your curiosity level up. it’s the stuff that drive is made of. don’t be discouraged! someday after you finish more courses in math, machine learning and artificial intelligence will be a piece of cake!