Credit card recognition app

Challenge

When building a face recognition app, we decided to validate the feasibility of creating our own character recognition solution. This Proof of Concept shows that we managed to train a machine-learning algorithm to recognize diacritical letters. There are many available solutions for card recognition yet their capabilities turned out insufficient. Having reviewed the existing solutions, we realized none of them met our requirements regarding Easy integration with iOS Seamless operation on iOS and Android Support for Polish characters Budget restrictions

Solution

Based on the image captured by the smartphone camera, the application analyses the card. A video frame is analysed by a Machine Learning model that works efficiently on iOS and Android systems both in online and offline mode. It’s capable of detecting and classifying objects such as digits and letters. The algorithm also uses probability to make the optimal decision based on the digit position. Only after one second, tens of frames are being analysed by our card recognition application. Based on the image captured by the smartphone camera, the application analyses the card. In order to optimise the area for analysis, we trained our app to look for a particular fragment on a card. The system analyses both the card number and the expiration date. We double-check the card number with Luhn algorithm. The app detects card providers (Visa, Mastercard, American Express) The app validates the expiry date.