Detecting Document Text with Amazon Textract - Amazon Textract

Detecting Document Text with Amazon Textract

To detect text in a document, you use the DetectDocumentText operation, and pass a document file as input. DetectDocumentText returns a JSON structure that contains lines and words of detected text, the location of the text in the document, and the relationships between detected text. For more information, see Detecting Text.

You can provide an input document as an image byte array (base64-encoded image bytes), or as an Amazon S3 object. In this procedure, you upload an image file to your S3 bucket and specify the file name.

To detect text in a document (API)

  1. If you haven't already:

    1. Create or update an IAM user with AmazonTextractFullAccess and AmazonS3ReadOnlyAccess permissions. For more information, see Step 1: Set Up an AWS Account and Create an IAM User.

    2. Install and configure the AWS CLI and the AWS SDKs. For more information, see Step 2: Set Up the AWS CLI and AWS SDKs.

  2. Upload a document to your S3 bucket.

    For instructions, see Uploading Objects into Amazon S3 in the Amazon Simple Storage Service Console User Guide.

  3. Use the following examples to call the DetectDocumentText operation.


    The following example code displays the document and boxes around lines of detected text.

    In the function main, replace the values of bucket and document with the names of the Amazon S3 bucket and document that you used in step 2.

    //Calls DetectDocumentText. //Loads document from S3 bucket. Displays the document and bounding boxes around detected lines/words of text. package com.amazonaws.samples; import java.awt.*; import java.awt.image.BufferedImage; import java.util.List; import javax.imageio.ImageIO; import javax.swing.*; import; import; import; import com.amazonaws.client.builder.AwsClientBuilder.EndpointConfiguration; import; import; import; import; import; import; import; import; import; import; public class DocumentText extends JPanel { private static final long serialVersionUID = 1L; BufferedImage image; DetectDocumentTextResult result; public DocumentText(DetectDocumentTextResult documentResult, BufferedImage bufImage) throws Exception { super(); result = documentResult; // Results of text detection. image = bufImage; // The image containing the document. } // Draws the image and text bounding box. public void paintComponent(Graphics g) { int height = image.getHeight(this); int width = image.getWidth(this); Graphics2D g2d = (Graphics2D) g; // Create a Java2D version of g. // Draw the image. g2d.drawImage(image, 0, 0, image.getWidth(this) , image.getHeight(this), this); // Iterate through blocks and display polygons around lines of detected text. List<Block> blocks = result.getBlocks(); for (Block block : blocks) { DisplayBlockInfo(block); if ((block.getBlockType()).equals("LINE")) { ShowPolygon(height, width, block.getGeometry().getPolygon(), g2d); /* ShowBoundingBox(height, width, block.getGeometry().getBoundingBox(), g2d); */ } else { // its a word, so just show vertical lines. ShowPolygonVerticals(height, width, block.getGeometry().getPolygon(), g2d); } } } // Show bounding box at supplied location. private void ShowBoundingBox(int imageHeight, int imageWidth, BoundingBox box, Graphics2D g2d) { float left = imageWidth * box.getLeft(); float top = imageHeight * box.getTop(); // Display bounding box. g2d.setColor(new Color(0, 212, 0)); g2d.drawRect(Math.round(left), Math.round(top), Math.round(imageWidth * box.getWidth()), Math.round(imageHeight * box.getHeight())); } // Shows polygon at supplied location private void ShowPolygon(int imageHeight, int imageWidth, List<Point> points, Graphics2D g2d) { g2d.setColor(new Color(0, 0, 0)); Polygon polygon = new Polygon(); // Construct polygon and display for (Point point : points) { polygon.addPoint((Math.round(point.getX() * imageWidth)), Math.round(point.getY() * imageHeight)); } g2d.drawPolygon(polygon); } // Draws only the vertical lines in the supplied polygon. private void ShowPolygonVerticals(int imageHeight, int imageWidth, List<Point> points, Graphics2D g2d) { g2d.setColor(new Color(0, 212, 0)); Object[] parry = points.toArray(); g2d.setStroke(new BasicStroke(2)); g2d.drawLine(Math.round(((Point) parry[0]).getX() * imageWidth), Math.round(((Point) parry[0]).getY() * imageHeight), Math.round(((Point) parry[3]).getX() * imageWidth), Math.round(((Point) parry[3]).getY() * imageHeight)); g2d.setColor(new Color(255, 0, 0)); g2d.drawLine(Math.round(((Point) parry[1]).getX() * imageWidth), Math.round(((Point) parry[1]).getY() * imageHeight), Math.round(((Point) parry[2]).getX() * imageWidth), Math.round(((Point) parry[2]).getY() * imageHeight)); } //Displays information from a block returned by text detection and text analysis private void DisplayBlockInfo(Block block) { System.out.println("Block Id : " + block.getId()); if (block.getText()!=null) System.out.println(" Detected text: " + block.getText()); System.out.println(" Type: " + block.getBlockType()); if (block.getBlockType().equals("PAGE") !=true) { System.out.println(" Confidence: " + block.getConfidence().toString()); } if(block.getBlockType().equals("CELL")) { System.out.println(" Cell information:"); System.out.println(" Column: " + block.getColumnIndex()); System.out.println(" Row: " + block.getRowIndex()); System.out.println(" Column span: " + block.getColumnSpan()); System.out.println(" Row span: " + block.getRowSpan()); } System.out.println(" Relationships"); List<Relationship> relationships=block.getRelationships(); if(relationships!=null) { for (Relationship relationship : relationships) { System.out.println(" Type: " + relationship.getType()); System.out.println(" IDs: " + relationship.getIds().toString()); } } else { System.out.println(" No related Blocks"); } System.out.println(" Geometry"); System.out.println(" Bounding Box: " + block.getGeometry().getBoundingBox().toString()); System.out.println(" Polygon: " + block.getGeometry().getPolygon().toString()); List<String> entityTypes = block.getEntityTypes(); System.out.println(" Entity Types"); if(entityTypes!=null) { for (String entityType : entityTypes) { System.out.println(" Entity Type: " + entityType); } } else { System.out.println(" No entity type"); } if(block.getPage()!=null) System.out.println(" Page: " + block.getPage()); System.out.println(); } public static void main(String arg[]) throws Exception { // The S3 bucket and document String document = ""; String bucket = ""; AmazonS3 s3client = AmazonS3ClientBuilder.standard() .withEndpointConfiguration( new EndpointConfiguration("","us-east-1")) .build(); // Get the document from S3 s3object = s3client.getObject(bucket, document); S3ObjectInputStream inputStream = s3object.getObjectContent(); BufferedImage image =; // Call DetectDocumentText EndpointConfiguration endpoint = new EndpointConfiguration( "", "us-east-1"); AmazonTextract client = AmazonTextractClientBuilder.standard() .withEndpointConfiguration(endpoint).build(); DetectDocumentTextRequest request = new DetectDocumentTextRequest() .withDocument(new Document().withS3Object(new S3Object().withName(document).withBucket(bucket))); DetectDocumentTextResult result = client.detectDocumentText(request); // Create frame and panel. JFrame frame = new JFrame("RotateImage"); frame.setDefaultCloseOperation(JFrame.EXIT_ON_CLOSE); DocumentText panel = new DocumentText(result, image); panel.setPreferredSize(new Dimension(image.getWidth() , image.getHeight() )); frame.setContentPane(panel); frame.pack(); frame.setVisible(true); } }

    This AWS CLI command displays the JSON output for the detect-document-text CLI operation.

    Replace the values of Bucket and Name with the names of the Amazon S3 bucket and document that you used in step 2.

    aws textract detect-document-text \ --document '{"S3Object":{"Bucket":"bucket","Name":"document"}}'

    The following example code displays the document and boxes around detected lines of text.

    In the function main, replace the values of bucket and document with the names of the Amazon S3 bucket and document that you used in step 2.

    #Detects text in a document stored in an S3 bucket. Display polygon box around text and angled text import boto3 import io from io import BytesIO import sys import psutil import time import math from PIL import Image, ImageDraw, ImageFont # Displays information about a block returned by text detection and text analysis def DisplayBlockInformation(block): print('Id: {}'.format(block['Id'])) if 'Text' in block: print(' Detected: ' + block['Text']) print(' Type: ' + block['BlockType']) if 'Confidence' in block: print(' Confidence: ' + "{:.2f}".format(block['Confidence']) + "%") if block['BlockType'] == 'CELL': print(" Cell information") print(" Column:" + str(block['ColumnIndex'])) print(" Row:" + str(block['RowIndex'])) print(" Column Span:" + str(block['ColumnSpan'])) print(" RowSpan:" + str(block['ColumnSpan'])) if 'Relationships' in block: print(' Relationships: {}'.format(block['Relationships'])) print(' Geometry: ') print(' Bounding Box: {}'.format(block['Geometry']['BoundingBox'])) print(' Polygon: {}'.format(block['Geometry']['Polygon'])) if block['BlockType'] == "KEY_VALUE_SET": print (' Entity Type: ' + block['EntityTypes'][0]) if 'Page' in block: print('Page: ' + block['Page']) print() def process_text_detection(bucket, document): #Get the document from S3 s3_connection = boto3.resource('s3') s3_object = s3_connection.Object(bucket,document) s3_response = s3_object.get() stream = io.BytesIO(s3_response['Body'].read()) # Detect text in the document client = boto3.client('textract') #process using image bytes #image_binary = stream.getvalue() #response = client.detect_document_text(Document={'Bytes': image_binary}) #process using S3 object response = client.detect_document_text( Document={'S3Object': {'Bucket': bucket, 'Name': document}}) #Get the text blocks blocks=response['Blocks'] width, height =image.size draw = ImageDraw.Draw(image) print ('Detected Document Text') # Create image showing bounding box/polygon the detected lines/text for block in blocks: print('Type: ' + block['BlockType']) if block['BlockType'] != 'PAGE': print('Detected: ' + block['Text']) print('Confidence: ' + "{:.2f}".format(block['Confidence']) + "%") print('Id: {}'.format(block['Id'])) if 'Relationships' in block: print('Relationships: {}'.format(block['Relationships'])) print('Bounding Box: {}'.format(block['Geometry']['BoundingBox'])) print('Polygon: {}'.format(block['Geometry']['Polygon'])) print() draw=ImageDraw.Draw(image) # Draw WORD - Green - start of word, red - end of word if block['BlockType'] == "WORD": draw.line([(width * block['Geometry']['Polygon'][0]['X'], height * block['Geometry']['Polygon'][0]['Y']), (width * block['Geometry']['Polygon'][3]['X'], height * block['Geometry']['Polygon'][3]['Y'])],fill='green', width=2) draw.line([(width * block['Geometry']['Polygon'][1]['X'], height * block['Geometry']['Polygon'][1]['Y']), (width * block['Geometry']['Polygon'][2]['X'], height * block['Geometry']['Polygon'][2]['Y'])], fill='red', width=2) # Draw box around entire LINE if block['BlockType'] == "LINE": points=[] for polygon in block['Geometry']['Polygon']: points.append((width * polygon['X'], height * polygon['Y'])) draw.polygon((points), outline='black') # Uncomment to draw bounding box #box=block['Geometry']['BoundingBox'] #left = width * box['Left'] #top = height * box['Top'] #draw.rectangle([left,top, left + (width * box['Width']), top +(height * box['Height'])],outline='black') # Display the image # display image for 10 seconds return len(blocks) def main(): bucket = '' document = '' block_count=process_text_detection(bucket,document) print("Blocks detected: " + str(block_count)) if __name__ == "__main__": main()
  4. Run the example. The Python and Java examples display the document image. A black box surrounds each line of detected text. A green vertical line is the start of a detected word. A red vertical line is the end of a detected word. The AWS CLI example displays only the JSON output for the DetectDocumentText operation.