test_ocr_local.py

"""
Local test script to verify OCR works on the sample CamScanner PDF.
Tests each page individually and prints the extracted text.
"""
import sys
import os
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))

import cv2
import numpy as np
from PIL import Image
import pdfplumber
import pytesseract

# Check if Tesseract is available
try:
    version = pytesseract.get_tesseract_version()
    print(f"Tesseract version: {version}")
except Exception as e:
    print(f"Tesseract not found: {e}")
    print("Trying to find tesseract...")
    # Try common Windows paths
    for path in [
        r"C:\Program Files\Tesseract-OCR\tesseract.exe",
        r"C:\Program Files (x86)\Tesseract-OCR\tesseract.exe",
        r"C:\Users\kmthu\AppData\Local\Programs\Tesseract-OCR\tesseract.exe",
    ]:
        if os.path.exists(path):
            pytesseract.pytesseract.tesseract_cmd = path
            print(f"Found tesseract at: {path}")
            break
    else:
        print("ERROR: Tesseract not found anywhere. Cannot test locally.")
        sys.exit(1)

pdf_path = r"c:\Users\kmthu\projects\puffnparse\sample file\CamScanner 25-03-2026 17.48 (1).pdf"

print(f"\n{'='*80}")
print(f"Testing PDF: {pdf_path}")
print(f"{'='*80}")

with pdfplumber.open(pdf_path) as pdf:
    print(f"Total pages: {len(pdf.pages)}")
    
    for page in pdf.pages:
        page_num = page.page_number
        print(f"\n{'='*80}")
        print(f"PAGE {page_num}")
        print(f"{'='*80}")
        
        # Convert page to image at 300 DPI (same as production code)
        page_image = page.to_image(resolution=300)
        pil_img = page_image.original
        
        # Convert PIL to OpenCV BGR
        open_cv_image = np.array(pil_img)
        if len(open_cv_image.shape) == 3 and open_cv_image.shape[2] == 3:
            img_bgr = cv2.cvtColor(open_cv_image, cv2.COLOR_RGB2BGR)
        elif len(open_cv_image.shape) == 3 and open_cv_image.shape[2] == 4:
            img_bgr = cv2.cvtColor(open_cv_image, cv2.COLOR_RGBA2BGR)
        else:
            img_bgr = cv2.cvtColor(open_cv_image, cv2.COLOR_GRAY2BGR)
        
        h, w = img_bgr.shape[:2]
        print(f"Image dimensions: {w}x{h} (landscape={w > h})")
        
        # Save original page image for inspection
        out_dir = r"c:\Users\kmthu\projects\puffnparse\sample file\test_output"
        os.makedirs(out_dir, exist_ok=True)
        cv2.imwrite(os.path.join(out_dir, f"page_{page_num}_original.png"), img_bgr)
        
        # Test fix_orientation
        from services.image_preprocessor import fix_orientation
        oriented = fix_orientation(img_bgr)
        cv2.imwrite(os.path.join(out_dir, f"page_{page_num}_oriented.png"), oriented)
        
        oh, ow = oriented.shape[:2]
        print(f"After orientation fix: {ow}x{oh}")
        
        # Run OCR on oriented image (simple grayscale, no heavy preprocessing)
        gray = cv2.cvtColor(oriented, cv2.COLOR_BGR2GRAY)
        text = pytesseract.image_to_string(gray, config='--psm 3').strip()
        
        # Show first 500 chars of output
        preview = text[:500] if len(text) > 500 else text
        print(f"\nExtracted text ({len(text)} chars):")
        print(f"---")
        print(preview)
        print(f"---")
        
        # Check if it looks like gibberish
        if text:
            words = text.split()
            long_words = [w for w in words if len(w) > 3]
            alpha_chars = sum(1 for c in text if c.isalpha())
            total_chars = len(text.replace(' ', '').replace('\n', ''))
            alpha_ratio = alpha_chars / total_chars if total_chars > 0 else 0
            
            print(f"\nQuality metrics:")
            print(f"  Total words: {len(words)}")
            print(f"  Words > 3 chars: {len(long_words)}")
            print(f"  Alpha ratio: {alpha_ratio:.2%}")
            print(f"  Likely gibberish: {alpha_ratio < 0.5 or len(long_words) < 5}")

print(f"\n{'='*80}")
print("Test complete! Check images in: " + out_dir)
print(f"{'='*80}")