Giles Bergel  0:03  
Hi everyone, I'm Giles Bergel, I'm based at Oxford and you can go to that site afterwards, I'm based in Visual Geometry, we're a general purpose computer vision group a pleace for research in computer vision and also applications in diverse fields. If you can turn to the next slide. This is our solution for image registration, it's very simple registration and visualisation tool. But it has quite a user-friendly interface. If you go to the next slide you can see, this is, I'm sorry I missed a slide, this is actually an application of it, this is the original application for our registration tool which is spotting variations in type settings, and you can see here, and Alan Sugar is modelling the so-called Wimbledon method of image comparison, which was extremely inefficient ing finding small differences in dense pages. If you could turn to the next slide. What you see here is a visualisation of the registration of these two images left and right as a simple flip registration; we also offer, red, green, similar to what Adam showed earlier, and various kinds of opacity. And we do this through feature extraction using what's now pretty well established called SIM feature extractions, which works pretty well for these materials. We are also fortunate in this case that  these materials are flat, but we do offer various transformations for materials that are variably curved. So that's the thin place lines for example for particularly this use case where there are books that are bound, and therefore, in varying states, and therefore the pages are variably curved in terms of (Next slide please). And we've been experimenting with applying it to other kinds of materials so here are some images of engraved materials on the left, and before and after images people really recognise conservation images ofthe Ghent alterpiece, you can see on the magnifying glass viewer, they're very much designed for a public audience. Next slide please. This is a more of a wish list, than a feature, than a roadmap. We would like to improve feature extraction perhaps using some of the methods, seen earlier, pre-registration polls, Possibly change detection, a certain set of our users, particularly the editors, are a bit reluctant has have change detection because they want to see things for themselves, they want to use this augmented, rather than artificial intelligence so be quite cautious I think that that feature. And of course, IIIF integration, the subject of this workshop is very high up on our feature list. Thanks very much.