Skip to main content

​
Overview

CoroNet uses Tesseract OCR as a fallback mechanism when the primary GPT-4o-mini vision model fails to detect a license plate. This ensures maximum reliability in license plate extraction.
Tesseract OCR is invoked automatically when the OpenAI model returns “NO_DETECTADA” (app.py:106-110). This dual-engine approach provides robust license plate detection.

​
How CoroNet Uses Tesseract

The application implements a two-tier OCR strategy:
  1. Primary: GPT-4o-mini vision model (app.py:44-80)
  2. Fallback: Tesseract OCR (app.py:106-110)
app.py
# Primary OCR with GPT-4o-mini
matricula = extract_plate_from_image(path)

# Fallback with pytesseract
if matricula == "NO_DETECTADA":
    image = Image.open(path)
    ocr_text = pytesseract.image_to_string(image, lang="eng")
    ocr_text = ocr_text.strip().replace(" ", "").replace("\n", "").upper()
    matricula = "".join([c for c in ocr_text if c.isalnum() or c == "-"])[:10]

​
Platform-Specific Installation

​
Windows Installation

1

Download Tesseract installer

Download the latest Tesseract OCR installer from the official repository:https://github.com/UB-Mannheim/tesseract/wikiChoose the appropriate version:
  • 64-bit: tesseract-ocr-w64-setup-5.x.x.exe (recommended)
  • 32-bit: tesseract-ocr-w32-setup-5.x.x.exe
2

Run the installer

  1. Double-click the downloaded .exe file
  2. Accept the license agreement
  3. Choose installation directory (default: C:\Program Files\Tesseract-OCR)
  4. Important: Select “Additional language data” and ensure English is selected
  5. Complete the installation wizard
3

Add Tesseract to PATH

This step is critical for pytesseract to find the Tesseract executable.
Option 1: During Installation
  • Check the box “Add Tesseract to PATH” in the installer
Option 2: Manual Configuration
  1. Open System Properties (Win + Pause/Break)
  2. Click “Advanced system settings”
  3. Click “Environment Variables”
  4. Under “System variables”, find and select “Path”
  5. Click “Edit” → “New”
  6. Add: C:\Program Files\Tesseract-OCR
  7. Click “OK” on all dialogs
4

Configure pytesseract path (if needed)

If Tesseract is not in your PATH, specify the path in your code:
import pytesseract
pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract.exe'
5

Verify installation

Open PowerShell or Command Prompt and run:
tesseract --version
Expected output:
tesseract v5.x.x
 leptonica-1.x.x

​
Windows Troubleshooting

Error: “pytesseract.pytesseract.TesseractNotFoundError”Add this to the top of app.py:
import pytesseract
pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract.exe'
Error: “Failed loading language ‘eng’”Re-run the installer and ensure English language data is selected.

​
Python Integration

CoroNet uses the pytesseract Python wrapper to interface with Tesseract OCR.

​
Install pytesseract

The pytesseract package is included in requirements.txt:34: 
pip install pytesseract
# or install all dependencies:
pip install -r requirements.txt

​
Basic Usage

import pytesseract
from PIL import Image

# Open an image
image = Image.open('license_plate.jpg')

# Extract text
text = pytesseract.image_to_string(image, lang='eng')
print(text)

​
CoroNet Implementation

The fallback OCR implementation in CoroNet (app.py:106-110): 
if matricula == "NO_DETECTADA":
    image = Image.open(path)
    ocr_text = pytesseract.image_to_string(image, lang="eng")
    ocr_text = ocr_text.strip().replace(" ", "").replace("\n", "").upper()
    matricula = "".join([c for c in ocr_text if c.isalnum() or c == "-"])[:10]
This implementation:
  1. Opens the image using PIL (Pillow)
  2. Runs Tesseract OCR with English language data
  3. Cleans the output (removes spaces, newlines, converts to uppercase)
  4. Filters to alphanumeric characters and hyphens
  5. Limits the result to 10 characters

​
Configuration Options

​
Custom Tesseract Path

If Tesseract is installed in a non-standard location, configure the path: 
import pytesseract

# Windows
pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract.exe'

# macOS (custom location)
pytesseract.pytesseract.tesseract_cmd = '/opt/homebrew/bin/tesseract'

# Linux (custom location)
pytesseract.pytesseract.tesseract_cmd = '/usr/local/bin/tesseract'

​
Language Configuration

Specify the language for OCR processing: 
# Single language
text = pytesseract.image_to_string(image, lang='eng')

# Multiple languages (English + Spanish)
text = pytesseract.image_to_string(image, lang='eng+spa')

​
Advanced OCR Configuration

For better license plate recognition: 
# Custom configuration for license plates
custom_config = r'--oem 3 --psm 7 -c tessedit_char_whitelist=ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789-'
text = pytesseract.image_to_string(image, lang='eng', config=custom_config)
Configuration options:
  • --oem 3: Use default OCR Engine Mode (LSTM)
  • --psm 7: Treat image as a single text line (good for license plates)
  • -c tessedit_char_whitelist: Restrict characters to alphanumeric + hyphen

​
Testing Tesseract

Test your Tesseract installation with a sample image: 
import pytesseract
from PIL import Image
import os

def test_tesseract():
    # Check if Tesseract is accessible
    try:
        version = pytesseract.get_tesseract_version()
        print(f"✓ Tesseract version: {version}")
    except Exception as e:
        print(f"✗ Tesseract not found: {e}")
        return
    
    # List available languages
    try:
        langs = pytesseract.get_languages()
        print(f"✓ Available languages: {', '.join(langs)}")
    except Exception as e:
        print(f"✗ Error listing languages: {e}")
    
    # Test with a sample image
    if os.path.exists('uploads/sample.jpg'):
        image = Image.open('uploads/sample.jpg')
        text = pytesseract.image_to_string(image, lang='eng')
        print(f"✓ OCR Result: {text}")

if __name__ == "__main__":
    test_tesseract()

​
Performance Considerations

Tesseract OCR is significantly faster than the OpenAI vision model but may have lower accuracy for complex images. The dual-engine approach balances accuracy and speed.

​
Typical Processing Times

  • OpenAI GPT-4o-mini: 2-5 seconds (network latency + API processing)
  • Tesseract OCR: 0.1-0.5 seconds (local processing)

​
When Tesseract is Used

Tesseract fallback is triggered when:
  1. OpenAI returns “NO_DETECTADA”
  2. OpenAI API is unavailable or rate-limited
  3. Image quality is too poor for vision model
  4. Network connectivity issues

​
Next Steps

Environment Variables

Configure API keys and application settings

OpenAI Setup

Set up OpenAI API for primary OCR

Build docs developers (and LLMs) love

Get started for freeTalk to us