Edge Devices
Edge devices have exploded with the advent of AI processors further pushing their usage due to a whole host of reasons. A few years ago, only large computer systems could use AI effectively and typically leveraged GPU’s in graphics cards to handle the low level yet abundant processing required to take input from one side of the neural net through weighted filters and provide a result. The problem apart from size of the equipment was the power needed to run the system over time.
Every edge ai computing device needs to run efficiently for their integration into a system along with needing to be isolated and work independently. Inline manufacturing monitoring systems need to be deployed on large production lines, in potentially small or remote areas of the factory. As a result, this means that the device may not be able to be connected to the internet or ethernet. When deploying AI, the isolation of the device has been a large challenge not just for factory processes but also automotive assisted driving modules.
Small form factor & Low Energy Solution
The advent of small low power consumption highly performant AI processors now allow a variety of integration methods for edge ai computing devices. You could mount the AI processor on the mainboard or use a modular card that can be put in any M.2 or mini-PCI socket and work with the mainboard processor to retrieve input information from sensors. However now there is the AI Box which essentially does the former in a pre-constructed case. The great news is that as the AI processor is passively cooled there is no need for fans or maintainable parts which can be useful in MIL-STD applications. Furthermore, all the company using one needs to do is decide what sensors to connect to the box.
The AI Box is small and optimized off the shelf to do exactly what you want it to without needing to mess around with assembling components. This is particularly useful in lower production runs or for SME’s that have either a lack of experience with hardware integration or facilities to conduct such operations. No hardware performance testing is required for the unit itself as everything has been designed and optimized already; this is something that shouldn’t be overlooked in the development process.
As the AI box utilizes an AI processor that comes with SDK integration tools it is very easy for a developer to integrate there existing AI solutions into the system, including container support for quicker product roll outs. If you don’t have an existing solution there are a whole multitude of out of the box AI solutions already for slight tweaking, saving developers a lot of time trying to get things working correctly.
A Quick Way to a Commercial Solution
AI Boxes are one box that can provide an unlimited set of solutions for the final product or service. For example, if you need a camera-based traffic monitoring system connect the camera to the box, select a predefined ‘out-of-the-box’ solution, make some tweaks for your requirement and you are good to go. As the system proves an SDK that abstracts the integration process, it is quick to setup with existing AI stacks. Have something already built in a container tested it but want to use something else in the box it is not a chore to do. These devices are so good that they can process up to 20 camera inputs in real-time. This means you can use them for practically every AI application that you can think of.
