Neural network analytics identify objects with high accuracy and enable video detection tools to work well in complex scenes with lots of visual noise. However, such analytics require sizable computing resources. This significantly decreases the number of cameras that can be connected to a server hosting AI analytics. It also makes the video surveillance system much more expensive.
The solution? Using AI accelerators and software to optimize the workload. It’s impossible to build a cost-effective video surveillance system that includes more than 10-20 AI analytics channels without them.
AxxonSoft supports products from the leading developers of these solutions.
OpenVINO™ is a toolkit for computer vision applications which extends workloads across Intel® hardware (including accelerators) and maximizes performance. The Intel® Distribution of OpenVINO™ toolkit is applied for neural network inference in AxxonSoft AI analytics tools. This makes it possible to significantly improve the performance of AxxonSoft neural network algorithms executed on standard CPUs and effectively extends the workload across all available Intel hardware.
AxxonSoft boosts AI performance even more using the latest Intel® Vision Accelerator Design products with Intel® Movidius™ VPU:
AxxonSoft neural network algorithms can run on the Huawei Atlas 300 AI accelerator card. Based on the Ascend 310 AI processor, the Atlas 300 AI card supports multiple types of data precision and is suitable for AI training and inference scenarios. The Atlas 300 AI accelerator card uses a groundbreaking Da Vinci architecture to unlock industry-leading performance and energy efficiency. It can be widely used for data centers and the intelligent edge.
AxxonSoft’s neural network video analytics work on NVIDIA’s CUDA-enabled graphics processing unit (GPU). AxxonSoft’s solution also supports NVIDIA’s GeForce 1050 Ti video cards, as well as newer ones with at least 2 GB of video RAM and that support Compute Capability 3.0 or higher.
CUDA is a parallel computing platform and programming model developed by NVIDIA. It enables dramatic increases in computing performance by harnessing the power of the graphics processing unit (GPU).