Scrapy with dynamic captcha

Question

I'm trying to break a captcha within a form from a website, but this captcha is dynamic, it doesn't have a URL instead it has something like this

src="captcha?accion=image"

What is the best option here? I have read something like using middlewares or something like that. Also I know it can be done with Selenium or Splash or another browser driver (screenshot), but i want to do it with just Scrapy, if it's possible of course.

Answer 1

Here's a complete solution to bypass the specified captcha using anticaptcha and PIL.

Due to the dynamic of this captcha, we need to grab a print screen of the img element containing the captcha. For that we use save_screenshot() and PIL to crop and save <img name="imagen"... to disk (captcha.png).
We then submit captcha.png to anti-captcha that will return the solution, i.e.:

from PIL import Image
from python_anticaptcha import AnticaptchaClient, ImageToTextTask
from selenium import webdriver

def get_captcha():
    captcha_fn = "captcha.png"
    element = driver.find_element_by_name("imagen") # element name containing the catcha image
    location = element.location
    size = element.size
    driver.save_screenshot("temp.png")

    x = location['x']
    y = location['y']
    w = size['width']
    h = size['height']
    width = x + w
    height = y + h

    im = Image.open('temp.png')
    im = im.crop((int(x), int(y), int(width), int(height)))
    im.save(captcha_fn)

    # request anti-captcha service to decode the captcha

    api_key = 'XXXXXXXXXXXXXXXXXXXXXXXXXX' # api key -> https://anti-captcha.com/
    captcha_fp = open(captcha_fn, 'rb')
    client = AnticaptchaClient(api_key)
    task = ImageToTextTask(captcha_fp)
    job = client.createTask(task)
    job.join()
    return job.get_captcha_text()

start_url = "YOU KNOW THE URL"
driver = webdriver.Chrome()
driver.get(start_url)
captcha = get_captcha()
print( captcha )

---

Output:

ifds

---

captcha.png

---

Notes:

• Use it at your own responsibility (be smart);
• You can improve the code by handling exceptions properly;
• anticaptcha is a paid service (0.5$/1000 imgs);
• I'm not affiliated with anticaptcha.