UC3M | Perfect images: how does AI fill in visual gaps?

---

There are numerous techniques that any user, whether experienced or not, can use to modify or alter an original image taken with any device, be it a professional camera or a cell phone camera. By applying filters on these images, the user makes use of internal algorithms that allow crossing the information transmitted by that image with a series of templates, which modify the original information. Techniques are also known to remove unwanted objects within images by drawing a mask over the area to be removed. Completing the image over that mask in a reasonable way is a technique that also makes use of artificial intelligence. But what happens if what we have is not an image on which a mask is applied, but a data set of fluid movements and obstructed areas from which we cannot extract information? That is the objective to investigate in this talk, where I will show different techniques that can be used and in which I will show you different results depending on whether the networks are trained with worse or better data. This activity is carried out in the framework of PERSEVERE, a scientific project funded by the European Union (EU) through a Marie Sklodowska-Curie Action (GA 101059984) of the Horizon Europe Framework Program. The views and opinions expressed are solely those of the author and do not necessarily reflect those of the EU. Neither the EU nor the granting authority can be held responsible for them. More information: https://persevere-msca.com/