Automated Document Classification using Deep Learning APIs

At Silversparro, we have delivered several large projects for Financial Institutions. One of the frequent requirements we have come across is Document classification – i.e. classifying images of documents such as PAN Cards, Aadhaar Card, and Driving License etc.

 Often, the requirement is variant of one of the following:

• Customers mail or upload documents to the website for availing financial services and have to classified before uploading to their system
• Customer submits a physical copy and back-office manually scans and uploads customer documents which need to be classified before uploading
• Older documents have to be scanned and digitally archived

This task is complicated as names of documents and content have no correlation. It’s common to find voter cards with file names such as ‘Img1’, ‘Scannedpdf’, and ‘KYCdoc’, ‘Aadhar’ or simply ‘sdssahsfbahjsfbashjf’. Also, many times a single page would have scanned images of two or more documents.

It was surprising to see this task happening manually at several financial institutions. Another solution being employed by financial institutions was that all such files were simply dumped into customer wise folder and extracted manually when needed. Getting such documents manually classification is not only time consuming but risky as the process risks leaking of personal information of customers.

Automatic classification is not a simple problem as documents can get fairly confusing and several documents like Driving License have huge variety of formats.We have trained a Deep Learning model to solve this. We use features from images, OCR of text and even file meta data to train a model. Model is hosted on a cloud GPU server and is accessible using APIs. Model inherently takes care of all possible rotations. This solution has been polished by testing at several large clients. Accuracy of the model is above 98%.

API is completely secure and fully automated. It doesn’t save any image or document on the disk – all images are processed on the RAM itself. API is end to end encrypted.

We are soft launching this general purpose Document Classification API for other financial companies. API is currently capable of classifying following categories. We can easily add more documents to this list upon request.

1.  PAN Card
2. Aadhaar Card
3. Passport
4. Driving License
5. ITR
6. Salary Slip
7. Voter ID
8. Bank Statements
9. Receipts
10. Board Certificate
11. Others

Checkout few sample results below (Intentionally Blurred for protecting privacy):

If you are interested in testing the API, just drop an email to my colleague aman.verma@silversparro.com for documentation and API endpoints. First 1000 API calls shall be absolutely free.

Silversparro Technologies is a Deep Learning Consulting Company working with large enterprises to solve their key business problems using expertise in Machine Learning and Deep Learning. Silversparro is working with clients across the world for computer vision, Natural Language and Large Transaction data use cases working for manufacturing, BFSI, healthcare verticals etc. Silversparro is backed by NVIDIA and marquee investors such as Anand Chandrashekaran (Facebook), Dinesh Agarwal (Indiamart), Rajesh Sawheny (Innerchef) etc.

Silversparro is founded by IIT Delhi Alumni – Abhinav Kumar Gupta, Ankit Agarwal and Ravikant Bhargava and is working for clients such as Viacom18, Policybazaar, Aditya Birla Finance Limited, UHV Technologies etc.