congruous 0.1.5

A python command-line tool to compare and generate accuracy reports for custom built OCR models.

Congruous

A python CLI that generates performance and accuracy reports of your OCR recognized data with the human - curated correct data

Installation

Works on python-3 environment and requires pip3 to install . In order to avoid environment conflicts setting up virtualenv is advisable.

$ pip install congruous

Just to confirm that congruous is installed correctly along with its dependency , run the --version comand.

$ congruous --version

This command should show you the lastest version number without any errors. However , its been reported that for some linux distribution python-tk is required to run this. Run the below command to fix it.

$ apt-get install python-tk

Quick Start

$ congruous --help

Lists all the available commands that can be used

Congruous mainly works based on the following three sub-commands ,

• store
• match
• report

Currently supported document types : pan , aadhar

Command : store

The purpose of store command is to handle all the database operations. The term store intends to store something or to keep hold of . Congruous uses an in-built sqlite database to handle the store operations .

Here are some of the options that can be used along with the store command.

--seed - feeds the database based on document type , prints number of records seeded if successful .

--head - displays the first 10 records that were seeded during the seed operation. can be used to verify if your matching against the right set of data.

--tail - displays the last 10 records that were seeded during the seed operation. can be used to verify if your matching against the right set of data.

--count - displays the total number of records seeded for a document type.

--drop - drops all the records for a particular document type that was seeded .

Command : match

This command performs the match operation and gives the performance of the OCR . Match can be performed either against seeded records or with another file using the options below.

--ocrd - mandatory field that passes custom built OCR parsed data as an input to the tool. By default it will be matched to seeded records

--hcd - to match with a another file instead of the seeded records in stores . passes a file to the tool

--report - generates a report on the the given current run of the congruous .

Command : report

This command is helpful to get insights about your OCR model performance over a period of time. Right now it just displays history - performance in the last few congruous runs . More options are to be included in the future versions

--history - displays the performance of you OCR for the last 10 runs.