Saturday, October 16th, 2021

Patent reveals Nintendo is working on upscaling technology

Image: Nintendo

Patents are pretty intriguing to keep track of in the tech industry all the time. Patents may represent ideas or products that never see the light of day, but they may also give clues for future releases.Newly released U.S. patent application The latter is possible because Nintendo supports technologies that can improve the visuals of the hardware.

Opened March 25, 2020 and released yesterday (September 30), the title of this application is “Machine Learning Image Conversion Systems and Methods”. The first language could be a head scratcher, but it’s essentially an idea similar to NVIDIA’s DLSS. This stands for Deep Learning Super Sampling for NVIDIA and works on some GPUs to upscale image resolution. When It guarantees real-time quality, yet very efficient and less burden on your graphics card. This is an impressive technology and has been at the center of much talk about how Nintendo can create new switch-style portable devices that deliver higher-resolution visuals while operating at lower power outputs. increase.

What makes this application interesting is that Nintendo is investigating it internally. The designated party to the application is Alexander de la Tre, co-founder of Nintendo Europe R & D. The patent “Introduction” also acknowledges that this is an area being investigated throughout the industry.

Machine learning allows a computer to “learn” a particular task without having to explicitly program the computer for that task. One type of machine learning system is called a convolutional neural network (CNN). This is a class of deep learning neural networks. Using such networks (and other forms of machine learning), for example, you can automatically recognize if a cat is in a photo. Learning is done by “training” the model with thousands or millions of pictures and recognizing when the cat is in the picture. While this is a powerful tool, the resulting processing of using (and training) a trained model can be computationally expensive even when deployed in a real-time environment.

Image up-conversion brings images generated at the first resolution (eg 540p resolution or 0.5 megapixel 960×540) to higher resolutions (eg 1080p resolution, 1920×1080, 2.1 megapixels). It is a technique that enables conversion. You can use this process to display the first resolution image on a higher resolution display. So, for example, you can display a 540p image on a 1080p TV and display it with higher graphic fidelity (depending on the nature of the up-conversion process) compared to displaying the 540p image directly in the traditional way (depending on the nature of the up-conversion process). For example, upscaling on a linear) 540 TV. Various image up-conversion techniques can offer a trade-off between speed (for example, the time it takes to convert a particular image) and the quality of the up-converted image. For example, if the up-conversion process runs in real time (for example, during a video game), the quality of the resulting up-converted image may be degraded.

Therefore, it will be understood that these areas of technology are in constant demand for new and improved technologies, systems and processes.

Ultimately, it’s not surprising that Nintendo is studying machine learning upscaling. In the future, if you choose to maintain switch-style form factors while increasing graphic fidelity, that’s likely to be an important factor. It’s also interesting to see if Nintendo will continue to use NVIDIA technology on future devices. If you are developing your own solution, you may not need NVIDIA’s DSSL solution.

Let us know your thoughts in the comments!