Abstract:

The paper presents the problem of quality assessment of image classifiers used in mobile phones for complimentary companion applications. The advantages of using this kind of applications have been described and a Narrator on Demand (NoD) functionality has been described as one of the examples, where the application plays an audio file related to a book page that is physically in front of the phone's camera. For such a NoD application, an image classifier is a key component. A thorough quantitative evaluation of the accuracy and robustness of such classifiers was conducted using 18 books of various sorts, 48 test classifiers and 10 testers. We analyzed the impact of the type and condition of the books that define the page classification problem, the position of the mobile device camera in the image acquisition process, the parameters of mobile devices used in tests, and selected options in a training process of classifiers.