Meet AI Edge
Not long ago the trend has changed from the Internet of Things to Artificial Intelligence. At UP, we didn't only stop at talking about this shift but we began to actually make it!
From the hype of the very first x86 Raspberry Pi-like UP Board to the most popular IoT development platform, UP Squared. Not forgetting the post-it sized computing platform, UP Core, we've been launching each exciting project successively.
With refueled excitement, we bring you UP AI Edge. The first embedded, ultra compact, high performance, low power consumption, Artificial Intelligence platform for Edge computing. Currently the AI Edge line has the following solutions.
UP Core Plus
- AI Plus
- Vision Plus
- Net Plus
- AI Core
- AI Core M2
Most IT solutions which are available today are focused on connecting Edge devices to the cloud and these deployments face challenges related to latency, network bandwidth, reliability and security. Experts in this field agree that not all the tasks and decision-making processes can be addressed in cloud-based models. UP AI Edge is the solution for those cloud limitations by bringing AI performance and hardware acceleration not ‘at’ but ‘ON’ the Edge of the Internet of Things and of the latest technology.
What is AI Edge?
The UP AI Edge platform, with its ultra-compact form factor, brings together three powerful Intel technologies: Intel Apollo Lake SoC, Intel® Movidius™ Myriad™ 2 VPU and Intel® Cyclone® 10GX FPGA. It is AI (Artificial Intelligence) built by IA (Intel Architecture) and comes together with a whole set of development tools.
UP AI Edge Stackable Platform
With the dual 100 pin board-to-board connectors, the UP Core Plus board is a powerful and expandable base to build up your solution. By stacking one of our carrier board solutions you can expand the functionality of the FPGA or VPU. Also with the open pin description (coming soon!) you can design your own carrier board.
UP AI Edge - base board and carrier boards
UP Core Plus
The credit-card sized (56 x 90 mm) board powered by a QuadCore Intel® Atom™ processor - formerly known as Apollo Lake-I.
This 64-bit architecture has an extremely high performance per watt ratio with a TDP starting at 6.5W (x5 – 3930). In this KickStarter you can select among three different CPU:
The board supports up to 8GB Dual Channel DDR4 2400 MHz, DP up to 4K, eDP, 2x USB 2.0, 1x USB 3.0, 1x USB 3.0 OTG, WiFi 802.11 AC 2T2R, 2 x CSI and it has 2 x 100 pin high speed connectors which bring to the carrier board I2C, I2S, ISH, UART, PCI-Express, USB 3.0 etc.
The credit-card sized (56 x 90 mm) companion board for UP Core Plus, powered by Intel® Cyclone® 10GX (105KLe-220KLe) with 1 GB DDR3, Display Port-IN connected to Cyclone 10gx for Video Acquisition, USB Type-C connector & GPIO connected to Intel® Cyclone® 10GX for data acquisition, LVDS-IN connected to Cyclone 10gx for Camera acquisition natively compatible to Basler LVDS cameras, 1 GBit Ethernet, 1 USB 3.0 and mini-PCIe connected to UP Core Plus via the 100 pin high speed connector.
The Intel® Cyclone® 10GX FPGA features include:
12.5 Gbps chip-to-chip transceiver I/O support and 6.6 Gbps backplane support
- High-performance 1,866 Mbps external memory interface
- 1.434 Gbps LVDS I/Os
- IEEE 754-compliant hard floating-point digital signal processing (DSP) blocks
Integrating 3x Intel® Movidius™ Myriad™ 2 VPU with 512 MB DDR, 1x GbLAN, 2x USB 3.0, and 1x mPCIEe for 3G/4G connectivity or other expansions in one ultra-compact board. It delivers three times the performance of a single UP AI Core module.
The credit-card sized companion board of UP Core Plus which features 4 Gigabit LAN powered by Intel I211AT ( option for I210 with support of TSN ), mPCIe to host either AI Core or 3G/4G modem, SATA, USB 3.0
UP AI Edge - AI Standard Expansions
The mini-PCIe module powered by Intel® Movidius™ Myriad™ 2 2450 with 512 MB DDR, the most advanced low-power technology for running deep learning and neural network on the Edge and processing video images.
If you only need the AI Core mPCI-E module, you can order it through up-shop.org.
AI Core M2
Integrating deep learning seamlessly into most edge computers such as the latest Intel NUC with AI Core M2.
Through the M.2 2280 B+M interface, accelerating the path to production from lab to field installation becomes a lot easier. The module is powered by 2 x Intel® Movidius™ Myriad™ 2 VPU with 512 MB DDR that is able to deliver double the performance when compared to a single VPU solution.
UP High-Definition Semi-industrial camera
With the AI Edge packages you will also find a AI Vision camera specifically designed for semi-industrial environment.
AI DEVELOPMENT KITUP Squared AI Vision Development Kit
This kit includes UP Squared E3950 CPU 4GB RAM, 64GB eMMC, UP HD camera, and AI Core module. Besides the hardware, it comes with a complete software package including Ubuntu 16.04.4 Desktop, OpenVINO™ toolkit, Open CL/ Movidius Driver, Arduino Create and Intel System Studio. It is a complete development kit with everything you need for AI vision development. The kit serves the purpose of bridging the gap between prototype and mass production and provides comprehensive cloud connectors such as Microsoft Azure, Amazon AWS, Google Cloud and IBM Bluemix. For more information click here.
Possible combination to unleash the power of your AI project
Besides using UP Core Plus as a single board computer, UP Core Plus is also a computer on module. All the carrier boards with the word "Plus" included in the naming is meant to work with UP Core Plus. Pick up the best configuration for your AI project and enable machine learning with FPGA or Networking. If you want, you can even have 5 Movidius™ Myriad™ 2 running all at the same time.
Enabling deep learning by multiple VPU
Are you trying to run more than one camera? Do you need a different network in VPU? Or do you require a whole different interface? UP AI Edge provides different options for the numbers of VPU and interface.
It goes along with most of the mainboards in the market for ease of your your development. Once the development is completed, we can also integrate the VPU into a custom single board computer.
Neural Compute Stick SDK
AI Core, AI Core M2 and Vision Plus are powered by Intel® Movidius™ VPU which allows the development of your software with Neural Compute Stick SDK.
This enables rapid prototyping, validation and deployment of Deep Neural Network (DNN) inference applications at the edge. Its low-power VPU architecture enables an entirely new segment of AI applications that are not reliant on a connection to the cloud. The AI Core and Vision Plus are fully compatible with Intel® Movidius™ Neural Compute SDK, allowing deep learning developers to profile, tune, and deploy Convolutional Neural Network (CNN) on low-power applications requiring real-time inferencing.
The NCSDK includes a set of software tools to compile, profile, and validate (check) DNNs as well as the Neural Compute API (NCAPI) for application development in C/C++ or Python. The NCSDK has two general usages: profiling, tuning, and compiling a DNN model on a development computer (host system) with the tools provided in the NCSDK.
Prototyping a user application on a development computer (host system), which accesses the neural compute device hardware to accelerate DNN inferences via the NCAPI. The following diagram shows the typical workflow for development with the NCSDK:
Open Visual Inference & Neural network Optimization (OpenVINO™ toolkit)
UP AI Edge is the official development kit for OpenVINO™ toolkit for smart cameras and video surveillance to robotics, autonomous vehicles, and more.
The Open Visual Inference & Neural network Optimization (OpenVINO™ toolkit) is a toolkit for developing applications and solutions that emulate human vision. Based on convolutional neural networks (CNN), the SDK extends workloads across Intel® hardware, and maximizes performance.
• Enables CNN-based deep learning inference on the edge
• Supports heterogeneous execution across OpenVINO™ toolkit accelerators—CPU, GPU, Intel® Movidius™ Neural Compute Stick, and FPGA—using a common API
• Speeds time-to-market via a library of functions and preoptimized kernels Includes optimized calls for OpenCV and OpenVX*
Open CL for AI Plus
The OpenCL standard is the first open, royalty-free, unified programming model for accelerating algorithms on heterogeneous systems.
Intel® FPGA SDK for OpenCL™ Overview
Intel® FPGA SDK for OpenCL™ is a world class development environment which enables software developers to accelerate their applications by targeting heterogeneous platforms with Intel CPUs and FPGAs.
This environment combines Intel’s state-of-the-art software development frameworks and compiler technology with the revolutionary, new Intel Quartus® Prime software to deliver next generation development environment that abstracts FPGA details while delivering optimized results. Intel FPGA SDK for OpenCL enables you to fully leverage the unique capabilities of FPGAs to deliver acceleration performance with power efficiency and low latency.
More information about Open CL here.
Create, Train & Deploy
From Lab to Field
Along with the development of the UP board, UP Squared and UP Core, we have developed an ecosystem in IoT, automation, retailing and robotics. This ecosystem includes partners from software, hardware and system integration. Together we have delivered a scalable solution in different continents.
The development of AI projects, however, needs one step more, i.e.: AI software partners. We have been working with some partners who are pioneers in deep learning. They develop algorithms according to different scenarios and deploy it into different target applications.
When you choose UP AI Edge, you get more than just a hardware platform. The whole ecosystem is at your service to aid you in achieving your goals.
Where we are at with this project?
UP Core Plus has reached the design verification test. This means that we already have the second batch of the engineering sample and it's under the qualification test now. So far, we see that there is a compatibility issue with NetPlus and WiFi/BT drivers. (picture or video of UP Core Plus)
AI Plus: As we launched this campaign, the first engineering sample has just reached the production line. We cannot wait to send them to our FPGA developers to verify it for us. (picture or video of UP Core Plus)
Vision Plus :We have finished the placement and will come to geber out soon. With our experience of AI Core, we believe that the first engineering sample expected in June will work just fine.
NetPlus: The first engineering sample is working well. We have also added AI Core on it to see how it works together.
Live-demo: UP Core Plus+ NetPlus+AI Core
AI Core M2: We have run the first engineering sample on an ASUS mother board. Indeed, 2 Intel® Movidius™ chips bring 2 times power in deep learning.
Live-demo: AI CORE M2 running on an Asus Motherboard
For full benchmarks please visit: https://openbenchmarking.org/result/1805177-KH-1802246FO66
About our Community, Shop, AAEON & Asus
UP is a brand owned by AAEON Europe, which is one of the ASUS group associated companies. In order to differentiate the market and fit in a different business model, we decided to have an UP garage team. The team is aiming to provide an embedded product, including hardware and software, to both professional makers and industrial markets.
Today we have more than 5000 users at UP Community and 7000 shoppers at UP Shop. All the products from this Kickstarter campaign will be available at UP Shop once the campaign ends. There is an UP Marketplace at UP Shop where we welcome everyone to sell their hardware, software or market ready solutions powered by UP.
Are you looking for a customization service? Yes, we can gladly assist you with that too! AAEON is an associate company owned by Asus, and we are ready to take care of your customization requirements.
Risks and challenges
AI Edge platform, with its ultra-compact form factor, brings together three powerful technologies of Intel: Intel Apollo Lake CPUs, Intel Movidius Myriad 2 and Intel Cyclone 10gx.
The hardware design has been very challenging thus far, because we integrated a series of multiple architecture on a credit-card sized board with multiple high speed busses of up to 12 Gbps for AI Plus (Cyclone 10Gx), DDR4 RAM at 2400 MHz, USB 3.0 bus, WiFi signal supporting wide bandwidth such 802.11 AC 2x2 (UP Core Plus).
Although we have already produced several working samples and the result so far is very promising, we are still in the design review phase for UP Core Plus, AI Plus and AI Vision to be sure that the final product will deliver performance, stability and robustness which is needed in industrial application.
Software validation will be a challenging factor as well and it will require the cooperation of different companies from our ecosystem.
Although the architecture we have selected come included with native support from different software development kits and tools, it’s always necessary to have a final internal and external validation of the product.
We are also working closely with Intel to make sure that the whole platform – including AI Plus powered by Cyclone 10gx – will be supported natively by the latest Intel Computer Vision SDK but we
need assistance from our community as well as your help to make this happen.
The big challenge at system level is to dissipate the CPU passively which has a TDP of up to 12W and the Cyclone 10gx which can reach 20W in its configuration with 220 KLe.
We may have to modify the housing unit partially in order to support active cooling, although this will be the last option to consider.
- (38 days)