Sony apparently does not want to rely solely on AMD’s proposal. The company is working on the technology of dynamic image upscaling in real time and according to the information presented – the corporation can take care of the quality of graphics in games on PS5 and the next-gen PS VR.
Sony has filed a new patent that sets out information about “Computer-implemented method of completing the image”. The Corporation explains in detail that the artwork may have some “Missing regions or corrupted data”and the technology being prepared does use machine learning to develop “masks” that will automatically replace fragments and hide them.
The technology is to learn how to automatically fill in the missing elements of the image – Deep Learning Super Sampling (DLSS) from NVIDIA works in a very similar way. As you know, Sony cannot use DLSS, and the alternative AMD FidelityFX Super Resolution collects mixed opinions.
According to the patent, the technology can be used in virtual reality because Sony recognizes well that VR needs “much more computing power” to display graphics that are satisfactory to the user. The idea is to provide players with better convenience and improve immersion.
Andy Bigos is one of the authors of the patent and according to his profile on LinkedIn, Sony’s senior chief engineer recently “Worked on the application of machine learning and neural rendering to graphics in real time on the PlayStation 5 console”which we can probably connect to the found project.