I am currently developing an AR app, and recently started using Model Targets. As the object we are enhancing is bigger than the augmentation itself (i.e. it is an industrial machine, where we cannot place markers, but it has several edges), we are generating MTs focusing on parts of its CAD model, and not its entirety.
To make the Model Target work, we have to point at the object so that the contours of the helper guide view match the edges of the part, and sometimes it is really hard.
We were wondering, if it could be possible to improve the Model Target recognition by making it detect the edges not only from a single point of view, but also from some of the nearby ones.