Log inGet started

Receipt OCR Python

Suggest Edits

The Python OCR SDK supports the receipt API for extracting data from receipts.

Using this sample below, we are going to illustrate how to extract the data that we want using the OCR SDK.

sample receipt

Quick Start

from mindee import Client, documents

# Init a new client
mindee_client = Client(api_key="my-api-key")

# Load a file from disk
input_doc = mindee_client.doc_from_path("/path/to/the/file.ext")

# Parse the document as an Invoice by passing the appropriate type
api_response = input_doc.parse(documents.TypeReceiptV4)

print(api_response.document)

Output:

----- Receipt V4 -----
Filename: receipt-with-tip.jpg
Total amount: 53.82
Total net: 40.48
Tip: 10.00
Date: 2014-07-07
Category: food
Subcategory: restaurant
Document type: EXPENSE RECEIPT
Time: 20:20
Supplier name: LOGANS
Taxes: 3.34 TAX
Total taxes: 3.34
Locale: en-US; en; US; USD;
----------------------

Receipt Data Structure

The receipt object JSON data structure consists of:

Document Level Prediction

For document level prediction, we construct the document class by combining the different pages in a single document.
This method used for creating a single receipt object from multiple pages relies on field confidence scores.

Basically, we iterate over each page, and for each field, we keep the one that has the highest probability.

For example, if you send a three-page receipt, document level will provide you one tax, one total, and so on.

# returns a unique object from class ReceiptV4
receipt_data.document

Page Level Prediction

We create the document class by iterating over each page one by one. Each page in the PDF is treated as a unique page.

For example, if you send a three-page receipt, page level prediction will provide you with three tax, three total and so on.

api_response.pages # [ReceiptV4, ReceiptV4 ...]

Raw HTTP Response

Contains the full Mindee API HTTP response object in JSON format

api_response.http_response # full HTTP request object

Extracted Fields

Each receipt object contains a set of different fields. Each field contains the four following attributes:

  • value (Str or Float depending on the field type): corresponds to the field value. Set to None if the >field was not extracted.
  • probability (Float): the confidence score of the field prediction.
  • bounding_box (Array[Float]): contains the relative vertices coordinates of the bounding box containing the >field in the image.
    If the field is not written, the bbox is an empty array.
  • reconstructed (Bool): True if the field was reconstructed using other fields.

Additional Attributes

Depending on the field type specified, additional attributes can be extracted in the receipt object.

Using the above receipt example, the following are the basic fields that can be extracted.

Document Type

  • document_type (string): Whether the document is an expense receipt or a credit card receipt.
document_type = api_response.document.document_type.value
print("document type: ", document_type)

Categories

  • category (string): Receipt category among predefined classes, as seen on the receipt.
category = api_response.document.category.value
print("purchase category: ", category)
  • subcategory (string): The receipt sub-category among predefined classes, as seen on the receipt.
subcategory = api_response.document.subcategory.value
print("purchase subcategory: ", subcategory)

Date

  • date (string): Payment date as seen on the receipt.
    • value (string): ISO 8601 date format (yyyy-mm-dd). European and imperial dates are both supported.
    • raw (string): In any format as seen on the receipt.
receipt_date = api_response.document.date.value
print("Date on receipt: ", receipt_date)

Locale

  • locale (string): Concatenation of language and country codes.
locale = api_response.document.locale.value
print("Locale code: ", locale)
  • locale.language (string): Language code in ISO 639-1 format as seen on the receipt.
language = api_response.document.locale.language
print("Language code: ", language)
  • locale.currency (string): Currency code in ISO 4217 format as seen on the receipt.
currency = api_response.receipt.locale.currency
print("Currency code: ", currency)
  • locale.country (string): Country code in ISO 3166-1 alpha-2 format as seen on the receipt.
country = api_response.document.locale.country
print("Country code: ", Country)

Orientation

  • orientation (number): The orientation field is only available at the page level as it describes whether the page image should be rotated to be upright.
    The rotation value is also conveniently available in the JSON response at: document > inference > pages [ ] > orientation > value.
    If the page requires rotation for correct display, the orientation field gives a prediction among these 3 possible outputs:
    • 0 degree: the page is already upright
    • 90 degrees: the page must be rotated clockwise to be upright
    • 270 degrees: the page must be rotated counterclockwise to be upright
orientation = api_response.document.orientation
print("Degree: ", orientation)

Supplier Information

  • supplier (string): Supplier name as written in the receipt.
supplier_name = api_response.document.supplier.value
print("Supplier Name: ", supplier_name)

Taxes

  • taxes (string): Contains tax fields as seen on the receipt.
    • value (float): The tax amount.
    • code (string): The tax code (HST, GST... for Canadian; City Tax, State tax for US, etc..).
    • rate (float): The tax rate.
    • basis (float): The amount used to calculate the tax.
taxes = api_response.document.taxes

# Loop on each Tax field
for tax in taxes:
   print(f"  tax amount: {tax.value}, tax_code: {tax.code}, tax_rate: {tax.rate}")

Time

  • time (string): Time of purchase as seen on the receipt
    • value (string): Time of purchase with 24 hours formatting (hh:mm).
    • raw (string): In any format as seen on the receipt.
time = api_response.document.time.value
print("Time: ", time)

Total Amounts

  • total_amount (number): Total amount including taxes
total_amount = api_response.document.total_amount.value
print("total with tax", total_amount)
  • total_net (number): Total amount paid excluding taxes
total_net = api_response.document.total_net.value
print("total without tax", total_net)
  • total_tax (number): Total tax value from tax lines
total_tax = api_response.document.total_tax.value
print("total tax", total_tax)

Tip

  • tip (number): Total amount of tip and gratuity.
tip = api_response.document.tip.value
print("Tip: ", supplier_name)

Questions?

Join our Slack

Updated 5 months ago