Hey fagots!

hey fagots!
looking for python AI libs that good with text recognition..
Im not here for arrogant answers like search for yourself, etc.
If you know a good document, share your knowledge
Is that what we are supposed to do right?

Attached: Screenshot from 2018-08-16 01-59-18.png (1920x1080, 678K)

Other urls found in this thread:


search for yourself, etc.

search for yourself

search for yourself
read the fucking manual
install gentoo


AhUqijsjsuzuzhzu UBUNTUUUuuU
iNSTALl gentoo

fuck gentoo why do i need gui for things i can do in binary distro

I've used nltk, gensim, and tensorflow.


First of all, what do you want to do with the text?

you dont need to worry about it fbi

> that good with text recognition
lmao wtf. What you need is a blockchain solution.

Come on, I can't recommend a tool without knowing the job.

The number of equal digits in my post isn't at all related to the number of letters in my agency's name

Assuming you're not a troll, you could just give a vague answer like "sentiment analysis" or "topic classification". I may even be able to recommend some specific algorithms depending on how detailed your response is.

That being said, I'm starting to feel like this is a bait thread.

Attached: 1506895451970.jpg (600x600, 18K)

its supposed to recognize characters from old scanned damaged documents

So you want just basic OCR or there are parts of the text missing and you want to infer?

There might be some easily available already trained models for OCR, but I don't know any from the top of my head

yes OCR self learning one for good results

>are parts of the text missing
It sounds like OP already answered that part of your question. OP is probably going to want to train a convolutional neural network to pull out the legible letters an then use an LSTM (or some other recurrent network) to infer the missing characters from the context.

You're probably going to want to look into tensorflow, keras, caffe, and pytorch. I've personally only ever used tensorflow and keras. They'd be decent for a job like this.

>trained models for OCR
Pre-trained models for OCR will be extremely helpful. However, OP could also train his own OCR model with EMNIST data.


What kind of text recognition? Please be clear with your questions if you're going to make a shitty thread, outside of the specified general for programming questions in the first place and then demand not to be told to look it up for yourself.

im starting with it i just ask if there is existing module or open source code that is build for that specific task

I just googled "tensorflow ocr" and github.com/emedvedev/attention-ocr came up. It looks somewhat similar to what you are trying to do and would probably perform reasonably OK on your task.

You may want to do some additional googling though. I spent all of 10 seconds searching on your behalf.


thank you

i can do it myself or make someone like you to do it for me thanks man

i use it its not perfect