1+ '''
2+ ML part for IP-APP-Price-Tracker project
3+ OCR invoice recognition using pytesseract module to extract:
4+ 1. Category of expense
5+ 2. Total Amount spent
6+ 3. Tax Amount
7+ 4. Score(accuracy of OCR)
8+ '''
9+
10+ #importing necessary modules
11+ from flask import Flask , request
12+ import pytesseract
13+ import os
14+ import cv2
15+ import json
16+ import re
17+ from Levenshtein import distance
18+ from werkzeug .utils import secure_filename
19+
20+ #local folder to save the image of invoice
21+ UPLOAD_FOLDER = 'D:/IP-App-Price-Tracker/src/uploads/'
22+ ALLOWED_EXTENSIONS = set ([ 'png' , 'jpg' , 'jpeg' ]) #the extensions allowed for images uplaoded
23+
24+ #starting Flask server to get image
25+ app = Flask (__name__ )
26+ app .config ['UPLOAD_FOLDER' ] = UPLOAD_FOLDER
27+
28+ '''
29+ OCR Processing Part
30+ '''
31+
32+ #basic dictionary to classify the text extracted from image after OCR processing
33+ dic = {'category' : [],
34+ 'amount' : [],
35+ 'score' : [],
36+ 'tax' : []}
37+
38+ #assigning directory for pytesseract
39+ pytesseract .pytesseract .tesseract_cmd = r'C:\Users\USER\AppData\Local\Tesseract-OCR\tesseract.exe'
40+
41+ #OCR processing the Image into a string
42+ def ocr_process (img , resolution = 450 , page_seg_method = '3' ):
43+ txt = ""
44+ txt = "" .join ([txt , pytesseract .image_to_string (img , lang = "eng" ,config = '--psm ' + page_seg_method )])
45+ return txt
46+
47+ #Categorising the Invoices according to some keywords from text extracted from the images
48+ def categories (result_string ,dic = dic ):
49+
50+ #Categories
51+ dining = re .findall ('(server)|(Food)|(Dining)|(order)|(table)|(restaurant)' ,result_string , re .IGNORECASE )
52+ apparel = re .findall ('(shirt)|(pant)|(jeans)|(clothing)|(sleeve)|(men)|(ladies)' ,result_string ,re .IGNORECASE )
53+ medicine = re .findall ('(medical)|(pharmacy)|(hospital)|(doctor)' ,result_string ,re .IGNORECASE )
54+ accessories = re .findall ('(accesories)|(earring)' ,result_string ,re .IGNORECASE )
55+
56+ #Appending the Categories into the dictionary
57+ if (len (dining )!= 0 ):
58+ dic ['category' ].append ('Dining' )
59+ elif (len (apparel )!= 0 ):
60+ dic ['category' ].append ('Apparel' )
61+ elif (len (medicine )!= 0 ):
62+ dic ['category' ].append ('Medical' )
63+ elif (len (accessories )!= 0 ):
64+ dic ['category' ].append ('Accessories' )
65+ return dic
66+
67+ #Scoring - removing frequently used words with '%d'
68+ def scoring (regex_expression , item ):
69+ item_revised = re .sub ('[0-9]{1,15}.{1,15}[0-9]{2}' , '%d' , item ).lower ()
70+ #removing most frequently repeated words
71+ item_revised = (item_revised .replace ('total' , '' )
72+ .replace ('amount' , '' )
73+ .replace ('balance' , '' )
74+ .replace ('due' , '' )
75+ .replace ('$' , '' )
76+ .replace ('₹' ,'' )
77+ .replace ('USD' , '' )
78+ .replace ('INR' ,'' )
79+ .strip ())
80+ #if all the strings are replaced with '%d' scpre=100, even if one extra character the score will become less
81+ score = 100 - distance ('%d' , item_revised )
82+ return score
83+
84+ #Extracting Total amount and Tax amount from the bill
85+ def amount_parsser (invoice_string , regex_expression , dic = dic ):
86+ #comparing using regular expression
87+ target_found = re .findall (regex_expression ['regex' ], invoice_string , re .IGNORECASE )
88+ tax_found = re .findall (regex_expression ['regex_tax' ],invoice_string , re .IGNORECASE )
89+
90+ if len (target_found ) == 0 :
91+ print ('Nothing matched' )
92+ return dic
93+ else :
94+ #Total Amount Processing
95+ for ind , item in enumerate (target_found ):
96+ if_tax_in_string = 'tax' in item .lower ()
97+ if_last_in_string = 'last' in item .lower ()
98+
99+ if any ([if_tax_in_string , if_last_in_string ]):
100+ del (target_found [ind ]) #delete the part with tax and last to find Total Amount
101+
102+ else :
103+ target_amount = re .search ('[0-9]{1,15}.{1,15}[0-9]{2}' , item )
104+
105+ if target_amount is not None :
106+ #extract the amount and score it at the same time
107+ score = scoring (regex_expression , item .lower ())
108+ amount = target_amount .group (0 ).replace (',' , '' )
109+
110+ #appending values into the dictionary
111+ dic ['amount' ].append (amount )
112+ dic ['score' ].append (score )
113+ #Tax Amount Processing
114+ for ind ,item in enumerate (tax_found ):
115+
116+ if_last_in_string = 'last' in item .lower ()
117+ if (if_last_in_string ):
118+ del (target_found [ind ])
119+
120+ tax_amount = re .search ('[0-9]{1,15}.{1,15}[0-9]{2}' , item )
121+
122+ if tax_amount is not None :
123+ #extract the amount and score it at the same time
124+ score = scoring (regex_expression , item .lower ())
125+ tax = tax_amount .group (0 ).replace (',' , '' )
126+
127+ #appending values into the dictionary
128+ dic ['tax' ].append (tax )
129+ return dic #return updated Dictionary
130+
131+ '''
132+ API PART
133+ '''
134+ #Checking allowed extensions of image
135+ def allowed_file (filename ):
136+ return '.' in filename and \
137+ filename .rsplit ('.' , 1 )[1 ].lower () in ALLOWED_EXTENSIONS
138+
139+ #Request file, process it and return json file with information
140+ @app .route ('/' , methods = ['GET' , 'POST' ])
141+ def upload_file ():
142+
143+ if request .method == 'POST' :
144+ # check if the post request has the file part
145+ if 'file' not in request .files :
146+ print ('No file part' )
147+
148+ file = request .files ['file' ]
149+ if file .filename == '' :
150+ print ('No selected file' )
151+
152+ #Getting the Image from Server and Calling the Processing Functions
153+ if file and allowed_file (file .filename ):
154+
155+ filename = secure_filename (file .filename )
156+ file .save (os .path .join (app .config ['UPLOAD_FOLDER' ], filename ))
157+ #reading the image
158+ img = cv2 .imread (os .path .join (app .config ['UPLOAD_FOLDER' ], filename ))
159+ img = cv2 .cvtColor (img , cv2 .COLOR_BGR2GRAY )
160+ #calling OCR function
161+ ocr_result = ocr_process (img )
162+ #Regular Expression for finding Total and Tax Amount
163+ totalAmountRegex = {'regex_tax' : '(gst[^0-9]{1,30}[0-9,]*\.\d\d)' ,
164+ 'regex' :'(?<!Tax )(?<!Sub)(?<!Sub )(Total[^0-9]{1,30}[0-9,]*\.\d\d)' }
165+
166+ #final dictionary that contains updated values
167+ result_dic = categories (ocr_result ,dic )
168+ result_dic = amount_parsser (ocr_result , totalAmountRegex , dic )
169+ #converting dictionary into json file
170+ app_json = json .dumps (result_dic , sort_keys = True )
171+ #returning json file back
172+ return app_json
173+
174+ if __name__ == '__main__' :
175+ app .run ()
0 commit comments