Sort:  

I really enjoyed reading this post!

I'm a bit curious... Have you noticed why those few examples were misinterpreted? Seriously bad writing, or something with the alignment or the line was broken...?

I'm glad that you like it:)

Well some of them had really bad writing but sometimes some other things can be factors like at this sample below:

greska.png

Because it has this black spot in the lower left corner and I was cutting the edges until it comes to first black pixel, it's recognized as 9 😢

It's nice to see that preprocessing was responsible and not the algorithm itself. With some more advanced method to replace the simplest thresholding, this becomes useful ;)

My dad was a dermatologist, so I've inherited some of his ineligible-writing skills :D It would be useful to have a machine to translate my writing for people!

a confusion matrix!

That's what I felt like reading some of the math!


Thanks for the mention. And always interesting to read a machine-learning post.

I love to believe that I actually write really nice and that I'm just too lazy to do it properly 😏

That's what I felt like reading some of the math!

Ohh I was right - I can bet that Gaussian formula scared you? :D

You're welcome, it's always interesting to write a machine learning post :)

This is sooooo cool!!! This is the coolest thing I have seen in a while, wow! Oh and, that formula is not the ugliest thing in your life and I am so going to use it in my post tomorrow about math and send you some people over, hihihi, I will even scare them with you :D

P.S. Do you have any idea how to fix the code to be more precise and maybe lower the error probability from 5% to maybe 2% or smaller?

Thanks Petra! 😊

Have you seen some uglier formulas? Now I'm curious.

You have definitely found the right person for scaring people with math, send them over! :D

Well, like I said to Alex above in comments - those mistakes appeared because some samples were really bad written and the algorithm by itself works nicely, but I assume that there can be some improvement.

Here's one idea - when cropping, we can after reaching first black pixels on image, to check few rows further to see do we have more black pixels - if not, we can consider it as a black spot and can continue cropping and preprocessing image. That way, we could eliminate that case of misinterpreted eight from above. I hope that I was clear :)

I see beauty in math so to answer your question, nope, I have not seen one ugly thing in math :)
Yeah, you were clear, I think... Like a loop part in the code that would repeat the cropping process after first checking if the pixel is a part of the number or not.

I see beauty in math

That's nice to hear :)

It is like seeing the Universe naked, kind of kinky but there are different levels of kinkiness, right? :)

Lmaoo, of course 😉

This is a very nice post, and the explanations are very clear. I am wondering how many people will try it at home. :)

Thank you @lemouth! :)

I would be very happy if someone would find himself in those things thanks to my blog, but I think that literally no one will try this at home :D

Maybe one day... who knows? :D

You have a very interesting article! It strike me coz I myself have a bad writing, I will always tell my parents, I'd rather type my report than printing it on the paper. But I think, for myself at least the reason is that I don't hold my pen correctly, many times I've heard people told me " I am holding my pen wrong", So I guess proper posture affect your handwriting.

Hi @nikolanikola!

Your post was upvoted by utopian.io in cooperation with @steemstem - supporting knowledge, innovation and technological advancement on the Steem Blockchain.

Contribute to Open Source with utopian.io

Learn how to contribute on our website and join the new open source economy.

Want to chat? Join the Utopian Community on Discord https://discord.gg/h52nFrV

i thought this post is best its increased my knowledge Good Job !

I would upvote your comment with at least 50$, only if I could ❤️

Coin Marketplace

STEEM 0.32
TRX 0.11
JST 0.034
BTC 66654.57
ETH 3250.95
USDT 1.00
SBD 4.33