P
P
petrushik2021-10-22 00:39:07
Python
petrushik, 2021-10-22 00:39:07

How to clearly recognize text from captcha?

There is such a captcha 6171dca0b9408954174612.png
you need to recognize the text with the condition. Raw Tesseract doesn't work, tried converting to B/W like this:

thresh = 35
            fn = lambda x: 255 if x > thresh else 0
            bw = primer.convert('L').point(fn, mode='1')

Better, but still very crooked, dabbled with the parameters of this code - to no avail.
I tried to drive RGB colors into the script so that everything except certain colors is deleted, but in each captcha each color changes its hue.
How can I implement reading text from such a captcha?

Answer the question

In order to leave comments, you need to log in

1 answer(s)
A
antares4045, 2021-10-22
@petrushik

61721e3575f7d636235081.png
Here is the neuron.
Only this particular network is trained in Chinese (I won’t even show how she read it). It's more of a proof of work.
Answer: CTPN for detecting words in a picture; CRNN for word recognition.
Relearn Russian and you will be happy.

Didn't find what you were looking for?

Ask your question

Ask a Question

731 491 924 answers to any question