On the use of pre-trained image classifiers for fingerprint-based indoor localization

Fingerprinting is a popular technique for indoor localization. It allows for the use of already existing infrastructure and offers satisfactory precision. The main drawback of fingerprinting is the arduous preparation time of gathering the fingerprint. Depending on the size of the building, this can be a big part of setting up an indoor localization system. In this paper, we will evaluate the precision of using pre-trained image lassifiers for our training and testing. First, we will turn the fingerprint into an image, then it is ready to be used for training. Although the use of pre-trained networks can save time, their use and accuracy are not comparable to custom architectures.