A team of researchers from Intel Labs presented a new approach that can greatly upgrade the look of synthetic images.
Stephan Richter, Hassan Abu AlHaija, and Vladlen Koltun from Intel Labs published a paper on an approach to enhancing the realism of synthetic images. The paper discussed how images can be enhanced by a convolutional network that uses intermediate representations produced by conventional rendering pipelines.
"The network is trained via a novel adversarial objective, which provides strong supervision at multiple perceptual levels," states the abstract. "We analyze scene layout distributions in commonly used datasets and find that they differ in important ways."
The paper explains a new strategy for sampling image patches during training. The approach also features multiple architectural improvements in the deep network modules used for photorealism enhancement.
What do you think about the approach? Imagine how this approach can help developers produce better-looking results without relying on traditional methods. You can find the paper here to learn more.