receiptparser 1.1

Receipt and bill parser using OCR

receiptparser

Summary

A receipt and bill parser written in Python. Can be used as a Python module or CLI tool.

It was originally based on receipt-parser, but has effectively been completely rewritten/replaced.

So far, only German receipts are supported, but other countries can be added using a simple YAML configuration file.

Recognition rate

To develop this tool, I used a set of 182 receipts in varying quality. Some of the were crumpled, most had been folded, etc. The result on this set of receipts is:

Total:             182
Company found:     171
Postal code found: 158
Date found:        159
Amount found:      114

If your receipts are sharp, uncrumpled, and have good contrast, I would expect a 97%-99% success rate, except for the total amount, which is harder to identify correctly. That may be closer to 75%.

Where applicable, I chose automation and quality over performance. For example, receiptparser scans every image twice, once unsharpened, and once sharpened, which raises the recognition rate around 6% but doubles the scan time.

Installation

Prerequisites

• Python 3
• PIP3
• tesseract

Install via PIP

pip3 install receiptparser

Install via Git

pip3 install -r requirements.txt
pip3 install .

Python usage

from receiptparser.config import read_config
from receiptparser.parser import process_receipt

config = read_config('my_config.yml')
receipt = process_receipt(config, "my_receipt.jpg", out_dir=None, verbosity=0)

print("Filename:   ", receipt.filename)
print("Company:    ", receipt.company)
print("Postal code:", receipt.postal)
print("Date:       ", receipt.date)
print("Amount:     ", receipt.sum)

CLI Usage

Examples

A simple example to read all images (.jpg) from a directory and print the recognized data to stdout:

receiptparser tests/data/germany/img/

You can customize the output as follows:

receiptparser -v0 --format "{date:%Y-%m-%d} - {company} - {postal} - {sum}.jpg" tests/data/germany/img/

In this case, -v0 suppresses any output, except for what you specify in the --format FORMAT parameter. FORMAT is a Python format string as specified here. The following values can be used in the format string:

• company: The recognized name of the company
• postal: The recognized postal code of the company
• date: The recognized date of the bill or receipt
• sum: The dollar (or Euro, or other currency) amount of the bill or receipt

Syntax

usage: receiptparser [-h] [-c CONFIG] [--config-file CONFIG_FILE] [-t TESSERACT] [-f FORMAT] [-v {0,1,2}] input

positional arguments:
  input                 file or directory from which images will be read

optional arguments:
  -h, --help            show this help message and exit
  -c CONFIG, --config CONFIG
                        built-in config to use
  --config-file CONFIG_FILE
                        like -c, but point to a file instead
  -t TESSERACT, --tesseract TESSERACT
                        output directory for OCR recognized text (default is to discard)
  -f FORMAT, --format FORMAT
                        format of the recognized output. default is pretty-printing
  -v {0,1,2}, --verbosity {0,1,2}
                        increase output verbosity