And again once a hoechst interesting papers from the field of artificial intelligence to which, inter alia, the Adobe Research Team participated. Who knows Photoshop, also should ever be tripped by the Content Aware Fill. This can missing image content from the program are supplemented conclusive. However, this does not always satisfactory because the Photoshop yes has no idea of u200bu200bthe imaged objects. So you're looking generally for similar patterns in the picture and tries to complement the filling bodies to procedurally. However, agent Deep Learnig can now be choosen a different approach. Has been seen millions of forming a trained network, so it can better guess what could be missing at the said location. And provided also yielded the first promtsignificantly better results than any method previously used. A nice example is this Leopard: PIC 1 Laesst to complement the various algorithms a missing segment in the middle of the picture, so succeed only to the neural network a plausible stuffing. (However, still missing here obviously the nose). But just when it comes in the future, unwanted objects from a (moving) to remove image assistants, may Neural Networks surely soon be playing the lead role. Especially since they can also combine well with popular form Procedure. A Difference Key should accrue with neuronal support example to Hoechst form. Let's wait what will come there so ...
more infos at bei arxiv.org;)
