How to Bypass Captchas with Python Selenium and OCR – A Step‑by‑Step Guide

1. Introduction

Hello, I am PiPi. Recently a fan asked about a Python web‑scraping captcha problem in a group, so I’m sharing the solution here.

Below is the original code:

from selenium import webdriver
from selenium.webdriver.common.by import By
import time
from PIL import Image
import ddddocr

ocr = ddddocr.DdddOcr()

options = webdriver.ChromeOptions()
options.add_argument('user-agent=Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/86.0.4240.198 Safari/537.36')
options.add_argument("--disable-blink-features=AutomationControlled")
driver = webdriver.Chrome(options=options)

# Open target page
driver.get('https://sol.sinosure.com.cn')
time.sleep(5)
driver.maximize_window()
# Locate captcha element and hover to load image
yanzhengma = driver.find_element(By.CSS_SELECTOR, '.pass-form-item.pass-form-item-code')
captcha_element = yanzhengma.find_element(By.CSS_SELECTOR, '.pass-label-img')
webdriver.ActionChains(driver).move_to_element(captcha_element).perform()
time.sleep(5)

# Get location and size of captcha element
location = captcha_element.location
size = captcha_element.size
print(location)
print(size)
# Capture full page screenshot
driver.save_screenshot('screenshot.png')

# Crop captcha from screenshot
left = int(location['x'])
top = int(location['y'])
right = int(location['x'] + size['width'])
bottom = int(location['y'] + size['height'])
captcha_screenshot = Image.open('screenshot.png').crop((left, top, right, bottom))
print(left)
print(top)
print(location)
print(bottom)
# Save cropped captcha and recognize
captcha_screenshot.save('captcha.png')
with open('captcha.png', 'rb') as f:
    img_bytes = f.read()
res = ocr.classification(img_bytes)
print('识别的验证码是：' + res)

The basic idea works, but the cropped area is slightly off, causing the OCR to miss the captcha.

Another snippet showing how to obtain the element's location and size:

location = captcha_element.location<br/>size = captcha_element.size

Some users found that the returned coordinates were near the password field rather than the captcha, even after trying to locate a larger container first.

2. Implementation

One contributor tried the code but encountered a driver‑version mismatch error, which is common when the local ChromeDriver does not match the installed Chrome browser.

The fix is to download the ChromeDriver version that matches your Chrome browser and ensure its folder is added to the system PATH.

Another approach suggested by a fan is to fetch the captcha image directly via its URL using requests, then pass the binary content to the OCR without saving a file:

This method avoids the cropping inaccuracies and works well for those less familiar with Selenium.

3. Conclusion

This article presented a Python web‑scraping solution for captchas, demonstrating both a Selenium‑based screenshot‑crop method and a direct request‑based approach, helping readers overcome common driver issues and recognition errors.