Neural Network and 4k Gaming!

Neural Network

A neural network is a series of algorithms that endeavors to recognize underlying relationships in a set of data through a process that mimics the way the human brain operates. In this sense, neural networks refer to systems of neurons, either organic or artificial in nature. Neural networks can adapt to changing input; so the network generates the best possible result without needing to redesign the output criteria.

Working

Artificial Neural Networks can be viewed as weighted directed graphs in which artificial neurons are nodes, and directed edges with weights are connections between neuron outputs and neuron inputs.

The Artificial Neural Network receives information from the external world in the form of pattern and image in vector form. These inputs are mathematically designated by the notation x(n) for n number of inputs.

Each input is multiplied by its corresponding weights. Weights are the information used by the neural network to solve a problem. Typically weight represents the strength of the interconnection between neurons inside the Neural Network.

The weighted inputs are all summed up inside the computing unit (artificial neuron). In case the weighted sum is zero, bias is added to make the output not- zero or to scale up the system response. Bias has the weight and input always equal to ‘1′.

The sum corresponds to any numerical value ranging from 0 to infinity. To limit the response to arrive at the desired value, the threshold value is set up. For this, the sum is passed through an activation function.

The activation function is set to the transfer function used to get the desired output. There are linear as well as the nonlinear activation function. Some of the commonly used activation function is — binary, sigmoidal (linear) and tan hyperbolic sigmoidal functions(nonlinear).

Industrial Use case!

Up sampling in games

For VR gaming to genuinely take-off, we need to make the experience significantly more life-like and sensible. This requires delivering the designs at extremely high-refresh rates and in any event in 4k resolution the virtual world closely resembles this present reality. Nonetheless, to do as such, we right now require amazingly top of the line and expensive equipment, which is the reason VR gaming has not gotten far and wide and mainstream.

So today, I need to share a new paper from Facebook reality labs that hopes to address this issue with the assistance of neural organizations. The paper is named “Neural Super-Sampling for Real-Time Rendering” by Xiao et. al. furthermore, it acquaints a profound learning strategy with up-sample the image delivered by a game engine continuously permitting us to keep up the high refresh rate too.

The main differentiation in this work compared to other up-sampling techniques is that this is specially designed to address up sampling in games while other approaches focus on photo realistic scenes.

This is significant in light of the fact that game renderings regularly contain visual ancient rarities like aliasing and spiked edges on 3D objects since they are misleadingly delivered by the game engine. Up sampling real-life images is an alternate undertaking since blurry pictures don’t contain the very kind of visual antiquities that are available in games.

It does as such by utilizing the 3D depth and motion data from the game engine which is the reason this technique yields much better outcomes that are likewise transiently consistent across successive frames.

This methodology likewise needs undeniably less computing power contrasted with straightforwardly delivering the game in high resolution, which I think will before long make it conceivable to make VR equipment substantially more affordable without forfeiting on nature of the experience.

Conclusion

We have presented a new method for neural up sampling of rendered video content. Our method achieves a new state of the art in super-resolving under sampled videos with extreme aliasing by using a new temporal up sampling design. Our method also compares numerically favorably to the existing state-of-the-art methods in super resolution, as well as in temporal super sampling used in game engines. We have also demonstrated the real-time performance of our method, which enables it to be used in real-time rendering applications in the future. We believe that the open design and the high-quality results reported in our method will also pave the road to a new body of work for neural super sampling in graphics.