Artificial intelligence and edge computing: a powerful combination for machine vision applications
13 Oct 2021 | Reading time: 3 min
How edge computing platforms can support AI applications
In recent years, we have seen exponential growth of data volumes due to the increasing number of data-generating digital devices. Cloud computing offers one way to handle these huge amounts of data: Data is transferred to the cloud, where it is processed, filtered and provided as decision-relevant information for industrial processes.
However, processing big data directly in the cloud involves long response times and high bandwidth requirements especially for image data. Edge computing can provide a viable solution to this dilemma. Let us show you how edge computing platforms can benefit industrial applications, especially those using artificial intelligence.
What is edge computing?
Edge computing allows data to be processed as close as possible to the source where it is generated, without interacting with the cloud. The term "edge" refers to the fact that data processing takes place at the edge of the network, i.e. in the network periphery, and as close as possible to the end device.
In this way, huge amounts of unstructured data can be analysed and filtered directly on the factory floor. Only relevant data is transferred to the cloud for long-term storage or distribution across multiple networks.
How can we benefit from edge computing?
Decentralised data processing at the "edge" of the network offers several advantages:
Significantly reduces response time and bandwidth requirements since data no longer needs to travel via the network to a remote data centre or to the cloud for processing.
Minimises data transfer and thus the possibility of data being stolen or corrupted by cyber criminals while being uploaded to the cloud.
Scalability becomes more flexible as computing capacity can be expanded through a combination of IoT devices and local "mini-data centres".
Reduced data volumes to be processed and stored in the cloud saves costs for data management.
Why should AI applications be moved to edge platforms?
Data privacy is one of the main reasons to move artificial intelligence to edge devices, especially when it comes to smartphones, smart speakers, home security cameras and robots.
Latency directly affects autonomous mobility capability in drones, robots and driverless cars, for which latency requirements are less than a millisecond.
Bandwidth is a critical factor for vision-based applications such as augmented reality (AR), virtual reality (VR) and mixed reality (MR). The wider the range of functions, the more bandwidth is required.
With edge computing, AI-based solutions can be deployed closer to data sources and the end customer. This enables next-generation smart services to be provided, whilst also placing high demands on hardware, driving the development of new platforms and chip types dedicated to these technologies.
Artificial intelligence on an edge platform with two GPUs
With the Neousys Nuvo-8208GC, STEMMER IMAGING provides the ideal solution for edge computing applications. The robust platform delivers impressive GPU power of up to 28 TFLOPS at FP32 for GPU-accelerated AI computing applications such as autonomous driving, vision inspection, security and surveillance. It is the first system of its kind to combine a superior CPU/GPU combination with significant computing power in a compact and ruggedised industrial-grade design.
In addition to the GPU-accelerated Edge AI platforms, STEMMER IMAGING offers other powerful solutions from Neousys Technology, an industry-leading provider of rugged embedded systems. A new generation of fanless embedded PCs equipped with the latest generation of processors and patented expansion cassette feature in the Nuvo series ensuring flexible use in demanding machine vision applications including a unique LP variant with a particularly flat housing profile.