> ## Content Index
> Fetch the complete content index at: https://www.aveshneeiyer.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# POC Prototype for Digital Account Opening
- URL: https://www.aveshneeiyer.com/document-classifier-id-drivers-license-passports/
- Published: 2021-07-01T17:39:44.000Z
- Updated: 2026-07-17T10:20:09.000Z
- Description: I have created a Prototype that classifies the 3 document types using Python, extracts text from the image using OCR; and eventually will be able to verify the barcode on ID, Passport & Drivers'.
- Author: Aveshnee iyer
- Tags: ocr, classifier, data science, microsoft ai, innovation, bitcoin, digital account opening

## Document Classifier (ID, Driver’s License, Passports)using a python model & OCR.

### Digital Account opening has been challenged with the capability to accept & verify documentation, capture and store metadata, amongst other legislative requirements.

**In this POC I have created a PROTOTYPE that:**

- classifies the 3 document types;
- extracts text from the image;
- and eventually will need to be able to verify the barcode.

**The key building blocks to create this end-to-end process are: document classification, text extraction, verification and storage.**

**Machine Learning models to classify documents types from images (ID, Passports, Driver's licenses) can be trained for the Personal market and Business customers.**

Nuances in the Business/ Wholesale market would be MoA, registration number as images and text extraction will include 'trading as', legal name, registration number, date of registration, country of registration.

###   

### **1\. Document Classification:** 

 **For ID's, Driver’s Licenses & passports.**

A classifier is a type of model that you can use to automate the identification and classification of a document type

**Model Performance Training - Precision and recall above 85%.**

![No alt text provided for this image](https://media-exp3.licdn.com/dms/image/C4D12AQHFM89GmwyW0w/article-inline_image-shrink_1000_1488/0/1625217323257?e=1631145600&v=beta&t=a2vwCA8l_Vx15Z6Xs1kzyNJ1LM5z_46EnmyprD4ie5M)

  
**Model Performance Prediction - results were generally above 94% accuracy in predicting the right document.**

![No alt text provided for this image](https://media-exp3.licdn.com/dms/image/C4D12AQHWtrkGxOkoSQ/article-inline_image-shrink_1000_1488/0/1625217578842?e=1631145600&v=beta&t=bU3JzEdVTQXP7BVYM0RQMXBIDyEe9yaMVNsUODQNIrU)

![No alt text provided for this image](https://media-exp3.licdn.com/dms/image/C4D12AQGaP_d3ofNpyA/article-inline_image-shrink_1000_1488/0/1625217711214?e=1631145600&v=beta&t=S5nvQZja2IGLVuHnsTTYEqz03ZYKa8eBp7e2BwT4Lis)

![No alt text provided for this image](https://media-exp3.licdn.com/dms/image/C4D12AQHgSb4vBb8GiQ/article-inline_image-shrink_1000_1488/0/1625217762008?e=1631145600&v=beta&t=QoIlW0ifrprkc0IiZCqiAVWTHFwKgn8x8eNSR8nxnts)

  
### **2\. Text Extraction:**

Extracting text from an image, in this scenario my ID:

![](https://storage.ghost.io/c/63/8c/638cc33b-5912-48b9-958c-94664f8e958a/content/images/2021/07/2021-07-02--3-.png)

### **3\. Verification:**

The barcode itself does not store information but points to the information stored in the database software. Using the code to create the barcode will establish if it is fraudulent or not. Can work for international ID’s as well as that info is available.

![](https://storage.ghost.io/c/63/8c/638cc33b-5912-48b9-958c-94664f8e958a/content/images/2021/07/barcode.jpg)

  
For more 

Follow me on Medium [https://medium.com/@aveshnee7](https://medium.com/@aveshnee7?ref=aveshneeiyer.com)