iSIZE, a specialist in deep learning for video delivery, has released first performance results taking full advantage of the special instructions enabled by Intel Advanced Matrix Extensions (Intel AMX), available on the 4th Gen Intel Xeon Scalable processor. The new Intel architecture means that AI-based video processing like iSIZE BitClear can be performed entirely within the CPU, eliminating the latency, energy and cost overheads of linking to a GPU or other external accelerator.
The volume of video to be processed, delivered and stored continues to grow exponentially. From archive footage and consumer-generated content to surveillance and drone images, much of this material is of limited resolution and marred by noise and distortion. Emergen Research predicts the market for security and surveillance video to grow by 18.7 percent by 2030, and drone cameras to grow at 26 percent over the same period. It has been reported that 500 hours of video are uploaded to YouTube every minute. The quality of video from such contexts is often hampered by dirt or rain on external lenses, drone vibration, storage and upload limits and degradation in archive footage. Therefore, to get the best results requires very sophisticated processing. iSIZE BitClear uses unique deep learning techniques to dynamically restore the video to the best possible quality, without affecting the perceptual intent of the content creator. It also permits AI-based video upscaling, allowing devices of limited native resolution to contribute high-resolution streams.
Intel AMX allows the advanced BitClear processing to be handled entirely within 4th Gen Intel Xeon Scalable processors. Measurements with BitClear have shown that 4th Gen Intel Xeon Scalable processors with Intel AMX deliver up to twice the performance of 3rd Gen Intel Xeon Scalable processors. This allows for BitClear to achieve real-time performance for denoising up to 1080p and denoising with upscaling for up to 2160p resolution, respectively. This was previously impossible without a specialised GPU or other accelerator hardware.
Sergio Grce, CEO and founder of iSIZE added “AI-based denoising and upscaling of video is well-known to require heavy use of GPU hardware. This tends to incur additional TCO in comparison to CPU-only execution. We have demonstrated our BitClear solution for AI-based video denoising running in real time on the 4th Gen Intel Xeon Scalable processors, which provides an excellent combination of performance and practicality.”