[Python] Seeking a robust Python scraper or library to bypass CAPTCHA on ecourts.gov.in

I’m trying to batch-download PDF judgments from https://judgments.ecourts.gov.in using plain requests, but their built-in math CAPTCHA keeps breaking my pipeline. I’ve rolled my own solution with easyocr to read the math challenge and a small parser to compute the result, but I still see:

• OCR mis-reads (e.g. confusing “×” vs “x”)
• Timeouts after solving multiple CAPTCHAs in quick succession
• Session-expiration errors that require me to reset the entire Session

What my CAPTCHA-solver code looks like


Below is the core function I use to fetch the CAPTCHA image, OCR it with EasyOCR, then evaluate the simple math expression. It often mis-extracts characters or returns malformed strings:

import re
import requests
import easyocr

ROOT = "https://judgments.ecourts.gov.in"
CAPTCHA_URL = ROOT + "/pdfsearch/vendor/securimage/securimage_show.php"

# Initialize the OCR reader only once
reader = easyocr.Reader(["en"], gpu=False)

def solve_captcha(session: requests.Session) -> str:
"""
1. Download the CAPTCHA image.
2. Pass raw bytes into easyocr for text extraction.
3. Parse the returned string as a simple arithmetic expression.
4. Return the computed result as a string.
"""
# Step 1: fetch image
res = session.get(CAPTCHA_URL, verify=False, timeout=30)
img_bytes = res.content

# Step 2: OCR the contents
ocr = reader.readtext(img_bytes)
if not ocr:
raise RuntimeError("OCR returned no text")

expr = ocr[0][1].strip() # e.g. "12 × 3"

# Step 3: parse "number operator number"
m = re.match(r"(\d+)\s*([+\-×Xx÷/])\s*(\d+)", expr)
if not m:
raise RuntimeError(f"Unexpected format from OCR: {expr}")

a_str, op, b_str = m.groups()
a, b = int(a_str), int(b_str)

# Step 4: evaluate
if op == "+": return str(a + b)
if op == "-": return str(a - b)
if op in ("×","X","x","*"): return str(a * b)
if op in ("÷","/"): return str(a // b)
raise RuntimeError(f"Unknown operator: {op}")

How I integrate it into my scraper


Right after solving the CAPTCHA, I post the answer to their token endpoint:

# after solve_captcha(session) returns a string `captcha_ans`:
token_resp = session.post(
CAPTCHA_TOKEN_URL,
headers={"X-Requested-With": "XMLHttpRequest", "Content-Type": "application/x-www-form-urlencoded"},
data={"captcha": captcha_ans, "ajax_req": "true"},
verify=False, timeout=30
)


If the token is accepted I proceed to request the PDF link; otherwise I get an error and retry.

The error I’m seeing


Even when the OCR reads something that looks valid, I sometimes get back:

400 Bad Request: errormsg=“Invalid CAPTCHA”


Or the session silently expires and I’m forced to call:

session = requests.Session()

What I’ve already tried

1. Resetting the Session every 25 downloads
2. Basic image preprocessing (grayscale → thresholding → dilation) before OCR
3. Fallback to Selenium + Tesseract (still unreliable)
4. Tuning EasyOCR parameters (GPU vs CPU, different model sizes)

What I’m looking for

1. Open-source projects or Python libraries with a more reliable CAPTCHA-bypass for ecourts.gov.in (or similar math/image CAPTCHAs).
2. Architectural patterns for a high-throughput scraper that gracefully handles CAPTCHA failures and session resets.
3. Recommendations for paid/hosted CAPTCHA-solving services or APIs that have a smooth requests-based integration (if no better open-source option exists).

Any pointers to existing GitHub repos, code samples, or best practices would be hugely appreciated—thanks!

Continue reading...