Assignment 03 – Course COMPUTER VISION II: APPLICATIONS (PYTHON)

 

Assignment 03 – Course COMPUTER VISION II: APPLICATIONS (PYTHON)

Performing OCR on Invoice Documents

Situation 

Mark works at an office where his job is to go through the invoices received and record the sender and the amount billed to a spreadsheet. It takes him only 1 minute to see the invoice and update the sheet. Still he finds it tedious and wants to automate this task. The main motivation for him is that he can then play on his system without any distractions. Will you help Mark in his pursuit of happiness?

Your Task

In this assignment, you will implement a OCR system, which looks at an image of the invoice and finds the following:

1.      The Billing Amount ( 15 Marks )

2.      Sender and Receiver Email IDs ( 15 Marks )

That's it!

It requires knowledge of OCR to get the data from the image and then some basic python skills to get the required values from the recognized text.

The assignment will be manually graded ( so that you cannot simply look at the image and simply do a print(billing_amount).

More information is given in the respective section.

Import Libraries.

import pytesseract

import keras_ocr

import matplotlib.pyplot as plt

import cv2

Read and display the Test Image

doc_img = cv2.imread('../resource/asnlib/publicdata/invoice.jpg', cv2.IMREAD_COLOR)

fig = plt.figure(figsize=(20, 10))

plt.imshow(doc_img[:,:,::-1])

plt.show()

Perform OCR

You need to write the code for performing OCR on the input image. You will have to apply the concepts learned in the previous sections to perform OCR on the above image.

After you perform the OCR - you have to parse the output so that you only print the email IDs and billing amount in $

# The next lists will store the outputs:

billing_amount = []

email_ids = []

 

###
### YOUR CODE HERE
###

# Pre-processing image for OCR, convert image to gray scale

doc_img_gray=cv2.cvtColor(doc_img,cv2.COLOR_BGR2GRAY)

 

# OCR with pytesseract

all_text = pytesseract.image_to_string(doc_img_gray)

 

# Parse  to get the required values from the recognized text

 

# The email adresses have the character '@' and the amount has the character '$'

 

# Split the OCR text in lines

 

all_lines = all_text.splitlines() # all_lines is a list with all lines, each one with several or none words

 

 

# We have to split each line in words

 

all_words=[]   # Empty list to sawe all the words

 

for line in all_lines:

    words = line.split()

    all_words += words

 

# Search for the words with special characters

for word in all_words:

    if(word.find('@')!=-1):

        email_ids.append(word)

    elif(word.find('$')!=-1):

        billing_amount.append(word)

 

# Print output