Matching and Painting using Gestalt Models

当艺术家在他们面前绘制或绘画真实世界时，他们不断地将他们所看到的与他们的艺术进行比较。他们对自己的艺术进行润饰，有时是为了使其更准确，有时是为了突出他们感兴趣的场景部分。关键是，艺术创作涉及到将某些东西看得比其他东西更重要，并且需要将艺术作品与现实世界相匹配的能力。如果我们用电脑代替艺术家，我们应该期待同样的事情是真实的，但目前情况并非如此。我们的建议是将照片和艺术品相匹配，这样机器就可以制作出更好的、新的艺术品。这样做比制作更好的艺术品有更多的好处，这是一种基本的技术，我们可以用它来帮助机器识别照片的重要部分，这有助于计算机通过人类的眼睛看到世界。

When an artist makes a drawing or painting of the real world in front of them, they constantly compare what they see to their art. They retouch their art, sometimes to make it more accurate, other times to bring out parts of the scene that interest them. The point is that making art involves seeing somethings as more important than others, and requires an ability to match artwork to the real world. If we replace the artist by a computer we should expect the same things to be true, but at the moment this is not the case. Our proposal is to match photographs and artwork so machines can make better - and new - art. Doing so offers more benefits than producing better artwork, it is a basic technique we can use to help machines identify important parts of a photograph, which helps computers see the world through human eyes just a little more.