I hope this blog helps those who are curious about what Super Resolution is and how AI algorithm can be accelerated and served through the data centers in On-premise or Public Cloud.
It also contains my desperate desire to reveal BLUEDOT's technology and product. 😉
(Original) (AI Upscaled)
Founded by experts with R&D and marketing experience in the field of video compression semiconductors in August 2019.
Experts in Semiconductor design, Artificial Intelligence, and Video Processing Algorithms are taking a challenge together under the slogan "fast and intelligent visual cloud computing".
Super Resolution converts Low-Resolution to High Resolution Video without losing details.
Super Resolution using Artificial Intelligence provides superior video quality compared to traditional upscaling, but requires a high amount of computation. Although GPUs are mainly used for training and inference of Artificial Intelligence Neural Networks, GPUs are often unsuitable for use in real-world services due to performance or cost reasons. Smaller neural networks can improve speed, but you may not get the desired results.
There are CPU, GPU, FPGA, and ASIC as computing devices. CPUs and GPUs have high versatility, but they have the disadvantage of being slower than FPGAs or ASICs that have built-in dedicated semiconductor circuits optimized for specific operations and functions, and consume very high power. ASICs cannot be changed when manufactured with specific circuits, so they are not suitable for use in cloud data centers that require processing of various types of workloads.