Running Tesseract OCR with pytesseract in AWS Lambda Using Container Images

Takahiro Iwasa

Jun 22, 2022

4 min read

Lambda Python

Developers can run Tesseract OCR with pytesseract using Lambda container images for efficient and scalable OCR operations.

For reference, you can pull the example code used in this blog post from my GitHub repository.

Ensure the following tools are installed on your system:

• AWS SAM
• Python 3.x

Organize your project as shown below:

/
|-- src/
|   |-- Dockerfile
|   |-- __init__.py
|   |-- app.py
|   |-- requirements.txt
|   `-- run-melos.pdf
|-- README.md
|-- __init__.py
|-- requirements.txt
`-- template.yaml

The following SAM template sets up the Lambda function triggered by EventBridge, since API Gateway has a maximum timeout limit of 29 seconds. The sample Python script execution exceeds this limit.

AWSTemplateFormatVersion: '2010-09-09'
Transform: AWS::Serverless-2016-10-31
Description: Tesseract OCR Sample with AWS Lambda Container Images using AWS SAM
Resources:
  TesseractOcrSample:
    Type: AWS::Serverless::Function
    Properties:
      Events:
        Schedule:
          Type: Schedule
          Properties:
            Enabled: true
            Schedule: cron(0 * * * ? *)
      MemorySize: 512
      PackageType: Image
      Timeout: 900
    Metadata:
      DockerTag: latest
      DockerContext: ./src/
      Dockerfile: Dockerfile

Create a Dockerfile to define the runtime environment. If your application processes text in a specific language like Japanese, set the LANG environment variable (line 3) accordingly to avoid encoding issues.

FROM public.ecr.aws/lambda/python:3.9

ENV LANG=ja_JP.UTF-8
WORKDIR ${LAMBDA_TASK_ROOT}
COPY app.py ./
COPY requirements.txt ./
COPY run-melos.pdf ./
RUN rpm -Uvh https://dl.fedoraproject.org/pub/epel/epel-release-latest-7.noarch.rpm \
    && yum update -y && yum install -y poppler-utils tesseract tesseract-langpack-jpn \
    && pip install -U pip && pip install -r requirements.txt --target "${LAMBDA_TASK_ROOT}"

CMD ["app.lambda_handler"]

Add the required libraries to requirements.txt.

pdf2image==1.16.0
pytesseract==0.3.9

The script converts a PDF to images, performs OCR, and logs the results.

import re
from datetime import datetime

import pdf2image
import pytesseract


def lambda_handler(event: dict, context: dict) -> None:
    start = datetime.now()
    result = ''

    images = to_images('run-melos.pdf', 1, 2)
    for image in images:
        result += to_string(image)
    result = normalize(result)

    end = datetime.now()
    duration = end.timestamp() - start.timestamp()

    print('----------------------------------------')
    print(f'Start: {start}')
    print(f'End: {end}')
    print(f'Duration: {int(duration)} seconds')
    print(f'Result: {result}')
    print('----------------------------------------')


def to_images(pdf_path: str, first_page: int = None, last_page: int = None) -> list:
    """ Convert a PDF to a PNG image.

    Args:
        pdf_path (str): PDF path
        first_page (int): First page starting 1 to be converted
        last_page (int): Last page to be converted

    Returns:
        list: List of image data
    """

    print(f'Convert a PDF ({pdf_path}) to a png...')
    images = pdf2image.convert_from_path(
        pdf_path=pdf_path,
        fmt='png',
        first_page=first_page,
        last_page=last_page,
    )
    print(f'A total of converted png images is {len(images)}.')
    return images


def to_string(image) -> str:
    """ OCR an image data.

    Args:
        image: Image data

    Returns:
        str: OCR processed characters
    """

    print(f'Extract characters from an image...')
    return pytesseract.image_to_string(image, lang='jpn')


def normalize(target: str) -> str:
    """ Normalize result text.

    Applying the following:
    - Remove newlines.
    - Remove spaces between Japanese characters.

    Args:
        target (str): Target text to be normalized

    Returns:
        str: Normalized text
    """

    result = re.sub('\n', '', target)
    result = re.sub('([あ-んア-ン一-鿐])\s+((?=[あ-んア-ン一-鿐]))', r'\1\2', result)
    return result

Run the following command to build the application:

sam build

Execute the following command to run the application locally:

sam local invoke

If an ECR repository does not exist, create one:

aws ecr create-repository --repository-name tesseract-ocr-lambda

Deploy the application:

sam deploy \
  --stack-name aws-lambda-tesseract-ocr-sample \
  --image-repository 123456789012.dkr.ecr.ap-northeast-1.amazonaws.com/tesseract-ocr-lambda \
  --capabilities CAPABILITY_IAM

After deployment, the Lambda function will run hourly and the OCR results will be written to CloudWatch Logs.

To clean up the provisioned AWS resources, use the following command:

sam delete --stack-name aws-lambda-tesseract-ocr-sample

Running Tesseract OCR in AWS Lambda using container images provides an efficient, scalable way to handle complex OCR workflows. With the flexibility of Docker, you can configure the environment to meet specific requirements.

Happy Coding! 🚀