Automatic Document Recognition
Project task
Creation of a program that recognizes specified types of documents (passport, driver’s license, vehicle registration certificate, etc.), as well as running the program on the customer’s servers.
Solution
A model has been created that supports multiple languages and recognizes a number of documents. The program is able to read data from documents and their copies in low-quality photo conditions.
Stages of work done:
- Analysis of existing approaches to document recognition
- Selection of the most promising approaches
- Training the model to recognize the required language
- Collecting a synthetic dataset for training the model
- Creating a model for the recognition of different types of documents
- Additional training of the model on the customer's data, metrics increment
- Implementation of the pre-trained model on the customer's servers
Personal data is completely protected
The program runs on the customer’s servers. All information and personal data does not go to third-party storage and is not transmitted anywhere.
Technology integrates into your system of work
The system is able to read data from a number of documents, but if the files need to be checked again, they are sent to an employee who will manually analyze the photo.
Business effect
The value of the project lies in automating routine procedures and eliminating the human factor from them. Such a solution saves time and money spent on hiring employees, as well as eliminates typos and errors in filling out data.
Areas of use
The case is relevant to organizations that serve mass customers in segments where it is necessary to store and verify data from customers' personal documents.
Project Team
Evgeny М.
CV Engineer
Mikhail V.
CV Engineer
Victor Sh.
CV Engineer
Dmitry I.
CV Engineer
Frank Sh.
Manager
