How to extract numbers from a complex captcha - python

I am trying to resolve captcha for the following image
!https://ibb.co/35X723J
I have tried using tessaract
data = br.open(captchaurl).read()
b = bytearray(data)
save = open(filename, 'wb')
save.write(data)
save.close()
ctext= pytesseract.image_to_string(Image.open(filename))

Here is a workaround. You need to clear a bit the image but you wont get a perfect result. Try the following:
try:
from PIL import Image
except ImportError:
import Image
import pytesseract
import cv2
file = 'sample.jpg'
img = cv2.imread(file, cv2.IMREAD_GRAYSCALE)
img = cv2.resize(img, None, fx=10, fy=10, interpolation=cv2.INTER_LINEAR)
img = cv2.medianBlur(img, 9)
th, img = cv2.threshold(img, 185, 255, cv2.THRESH_BINARY)
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (4,8))
img = cv2.morphologyEx(img, cv2.MORPH_CLOSE, kernel)
cv2.imwrite("sample2.jpg", img)
file = 'sample2.jpg'
text = pytesseract.image_to_string(file)
print(''.join(x for x in text if x.isdigit()))

Option 1:
I think using Pytesseract should solve the issue. I tried out your code and it gave me the following result when i gave in the exact cropped captcha image as input into pytesseract:
Input Image:
Output:
print(ctext)
'436359 oS'
I suggest you don't give the full page url as input into pytesseract. Instead give the exact image url as "https://i.ibb.co/RGn9fF5/Jpeg-Image-CS2.jpg" which will take in only the image.
And regarding the extra 'oS' characters in the output, you can do a string manipulation to chop off the characters other than numbers in the output.
re.sub("[^0-9]", "", ctext)
Option 2:
You can also use google's OCR to accomplish this which gives you the exact result without errors. Though I have shown you the web interface of it, google has nice python libraries through which you can accomplish this using python itself. Looks like this:

Related

How to read image from URL (urllib.error.HTTPError: HTTP Error 403: Forbidden)

I have a code that returns an image (API) :
url1 = "https://api.twilio.com/2010-04-01/Accounts/AC9567862a70d7b000488e8ba80bc19787/Messages/MMebd9e089637c2c740608199af762fc19/Media/ME49a13b9e0732d78b84fe81d6c3c5a3f2"
Manually open the image on broswer automatically converts the https address to below:
url2 = "https://s3-external-1.amazonaws.com/media.twiliocdn.com/AC9567862a70d7b000488e8ba80bc19787/f28ad1174d0871048be68207a88d5cea"
from skimage import io
img = io.imread(url1) # < ---- error here
pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract.exe'
text = pytesseract.image_to_string(img)
I'm trying to read the image from the url1 and perform OCR on the image.
(That is to read the image directly from (url1) without the need to open it on broswer and use the alternate address)
However, io.imread(url1) gives the error "urllib.error.HTTPError: HTTP Error 403: Forbidden".
io.imread(url2) does not return any error.
Is there any way to read the image directly from url1 ?
You can use this code in the second line as skimage.io.imread() takes URL as a string input.
img = io.imread(String(URL))
Also you can use cv2 module instead.
# importing OpenCV(cv2) module
import cv2
img = cv2.imread(URL)
# Output img with window name as 'image'
cv2.imshow('image', img)

Tesseract doesn't recognize characters(numbers)

I'm trying to read a water level from a photo file(jpg or png).
but Tesseract does not read any of character at all even I cut the all the unnecessary area of photo.
I put a print function at the end of the code to see recognized number, but nothing appeared.
enter image description here
Here is the photo I have and under neath is the Python code.
import cv2
import os
from PIL import Image
import pytesseract
pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract'
image = cv2.imread ("water002.jpg")
#gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
gray = cv2.Canny(image, 300, 350)
filename = "{}.png".format(os.getpid())
cv2.imwrite(filename, gray)
text = pytesseract.image_to_string(Image.open(filename), lang=None)
#os.remove(filename)
print(text)
cv2.imshow("Image", image)
cv2.waitKey(0)
Is there any tip to let Tesseract to read the level or any other method?

I am getting different results, If I read image by base64 vs read from storage

I have a python flask application. I created an OCR API.
I am getting correct results, If I read the image from storage like this:
img = cv2.imread('test1.jpeg')
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
But I am getting a bit incorrect results, when I read the same image using Base64. Like this
img = cv2.imdecode(np.frombuffer(request.files['fileToUpload'].read(), np.uint8), cv2.IMREAD_COLOR)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

Capture a web in image with OpenCV

I want to do a practice that consists of capturing webs in jpg, but it did not just work (I am newbie), this is the code I use.
import numpy as np
import urllib
import cv2
def url_to_image("http://www.hereiputweb.com"):
resp = urllib.urlopen("http://www.hereiputweb.com")
image = np.asarray(bytearray(resp.read()), dtype="uint8")
image = cv2.imdecode(image, cv2.IMREAD_COLOR)
return image
The code I got it from a manual but it gives me fault in the line:
def url_to_image("http://www.hereiputweb.com"):
I think I indicated the web incorrectly, very far I should not be .. tried several forms but nothing .. what do I do wrong?
regards
There is a really brief tutorial (https://docs.python.org/3/tutorial/).
The relevant part would be https://docs.python.org/3/tutorial/controlflow.html#defining-functions
So, you should define your function as follows:
def url_to_image(url):
resp = urllib.urlopen(url)
image = np.asarray(bytearray(resp.read()), dtype="uint8")
image = cv2.imdecode(image, cv2.IMREAD_COLOR)
return image
I have not checked the implementation works ;)
Then you can use your function:
url = "http://www.hereiputweb.com"
my_image = url_to_image(url)
The problem is not with your implementation, it's with your URL!
This method require a functioning URL that returns an image. The URL you're using is not an image.
Try using an URL of an image (e.g: some URLs that end with .jpg) and it shall work!
Remember that the URL must be on-line!

opencv python face detection from url images

I have no problem getting the opencv face detection using haar feature based cascades working on saved images:
from PIL import Image
face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')
eye_cascade = cv2.CascadeClassifier('haarcascade_eye.xml')
img = cv2.imread('pic.png')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray, 1.3, 5)
but I can't figure out how to open a url image and pass it into face_cascade. I've been playing around with cStringIO, but I don't know what to do with it...
import cv2.cv as cv
import urllib, cStringIO
img = 'http://scontent-b.cdninstagram.com/hphotos-prn/t51.2885-15/10424498_582114441904402_1105042543_n.png'
file = cStringIO.StringIO(urllib.urlopen(img).read())
source = Image.open(file).convert("RGB")
bitmap = cv.CreateImageHeader(source.size, cv.IPL_DEPTH_8U, 3)
cv.SetData(bitmap, source.tostring())
cv.CvtColor(bitmap, bitmap, cv.CV_RGB2BGR)
is it possible to work with a numpy array instead?
source2 = Image.open(file)
imarr=numpy.array(source2,dtype=numpy.uint8)
I'm a beginner, so I apologize for the poor explanation.
thanks a lot in advance!!
In your first example you are using OpenCV2.imread to read your image in the second you are presumably using PIL.Image then trying to convert.
Why not simply save the file to a temp directory and then use OpenCV2.imread again?
Or in another way you can use VideoCapture() class to open url image.
See the C++ code below,
VideoCapture cap;
if(!cap.open("http://docs.opencv.org/trunk/_downloads/opencv-logo.png")){
cout<<"Cannot open image"<<endl;
return -1;
}
Mat src;
cap>>src;
imshow("src",src);
waitKey();

Categories

Resources