Project
The possibility confirmed for technical implementation of the real-time text recognition
Goal
To check whether it is technologically feasible to develop an application that allows recognizing the text and converting it into a readable form.
Obstacle
The application used a third-party Tesseract framework. There was no possibility to influence the recognition accuracy, since it depended on many factors such as the quality of the original image, lighting, font, and shooting angle.
Solution
To confirm the concept, two different approaches were integrated within the same application. On the one hand, we used Vision and MLCore which are the native iOS components. And on the other hand, we used the third-party Tesseract approach.
Vision determines the presence of characters in each video frame. Tesseract recognizes the text in a specific video frame.
We worked with the tools as consumers only, since the key goal at that time was to demonstrate options for recognizing visual images, as well as to analyze the capabilities of iOS 11 in terms of Machine Learning.
The application was ready in 2 hours. The client received confirmation that such a feature can be implemented and used within more complex applications.
