Cardoso, Jorge C. S.Belo, AndréSablatnig, Robert and Wimmer, Michael2018-11-112018-11-112018978-3-03868-057-42312-6124https://doi.org/10.2312/gch.20181348https://diglib.eg.org:443/handle/10.2312/gch20181348Augmented Reality (AR) development frameworks have different recognition performance on different kinds of target images. In this work, we studied AR frameworks applied in the context of outdoor Roman mosaic ruins with the final aim of developing a multi-platform mobile AR application. We started by analysing the documented features of existing frameworks to determine the feasible ones. This resulted in the selection of three frameworks: CraftAR, PixLive, and Wikitude. We then experimentally evaluated the performance of the target recognition features against real mosaics by measuring the recognition delay, minimum required target area, visual alignment and visual stability. Results indicate a good recognition rate for CraftAR and a poor recognition rate for Wikitude. CraftAR showed better recognition delay and visual stability, while PixLive showed better minimum required target area and visual alignment.Humancentered computingMixed / augmented realityEmpirical studies in ubiquitous and mobile computingEmpirical studies in HCISmartphonesComputer graphicsMixed / augmented reality10.2312/gch.20181348119-128