Aaeon has launched an “UP AI Edge” family of products that builds on a new Apollo Lake based “UP Core Plus” SBC with stacking AI companion boards based on the Movidius Myriad 2 or Intel Cyclone 10GX plus add-ons including a quad-GbE board and a camera.

Aaeon Europe quickly met its modest $11K Kickstarter goal for the new UP AI Edge ecosystem, which builds on its UP board products and community. The centerpiece is a new UP Core Plus SBC, although the official, Ubuntu-equipped UP AI Edge development package uses the larger, more feature-rich UP Squared SBC.

UP Core Plus SBC with Net Plus underneath and heatsink on top (left) and UP Core Plus chassis

There are 22 packages overall, most of which combine the UP Core Plus with one of three major boards that stack on top of the UP Core Plus. The AI Plus board incorporates an Intel Cyclone 10 GX FPGA, which we also saw pop up last week on Arduino’s new MKR Vidor 4000 board. As an alternative you can go with a Vision Plus board that includes three of Intel’s Movidius Myriad 2 visual processing units (VPUs) for deep learning processing.


UP Core Plus, front and back
(click image to enlarge)

There’s also a Net Plus board with 4x GbE ports that includes a mini-PCIe slot. Three of the packages combine the UP Core Plus, Net Plus, and UP AI Core, a mini-PCIe form factor board with a Myriad 2 chip that was announced in March as an AI companion to the UP Squared. Another three packages combine the AI Core with the UP Core Plus SBC and the mini-PCIe equipped AI Plus FPGA board.


UP AI Core, front and back
(click images to enlarge)

Finally, there are packages that lack an UP SBC, but instead supply a new AI Core M2 module with one or two Myrad 2 chips. Instead of using mini-PCIe like the single-Myriad 2 equipped AI Core, the AI Core M2 has an M.2 2280 B+M form factor and interface. There’s also an optional USB Vision camera with up to HD resolution, an OV2735 sensor, and an extended temperature range.


AI Core M2, front and back (left) and USB Vision camera

The only mention of OS support is the Ubuntu 16.04 image that ships with the UP Squared AI Vision Kit, which is also the official development kits for the UP AI Edge family. The kit includes the high-end, Apollo Lake based UP Squared SBC 4GB RAM, 64GB eMMC, a power supply with EU or US power cords, a chassis, and the AI Core mini-PCIe module.

UP Squared

The Ubuntu stack also includes Intel’s OpenVINO computer vision toolkit, which is optimized for the Myriad 2. Also available is the Arduino Create development environent, an Open CL/ Movidius Driver, Intel System Studio, and cloud connectors for Microsoft Azure, Amazon AWS, Google Cloud, and IBM Bluemix.

You can also use the Neural Compute Stick SDK available for the Myriad 2 equipped Intel Movidius Neural Compute Stick for “rapid prototyping, validation and deployment of Deep Neural Network (DNN) inference applications at the edge,” says Aaeon. Finally, the dev kit provides the Intel FPGA SDK for OpenCL.

UP AI Edge Kickstarter packages
(click images to enlarge)

The AI Core M2 based Kickstarter packages are available in July, and the UP Squared development bundle ships in August. All the packages with the UP Core Plus board ship in October. The UP Core Plus board with power supply and chassis starts at 159 Euros ($186), and the combo packages range from 319 Euros ($374) up to 679 Euros ($795) for a high-end UP Core Plus with AI Plus, AI Core, and the camera (see the feature and price charts above and below).


More UP AI Edge packages (left) and UP Core Plus with Net Plus
(click images to enlarge)

Here’s a closer at the Plus boards in the UP AI Edge ecosystem.

 
UP Core Plus

The 90 x 56mm UP Core Plus board is larger than the 66 x 56.5mm UP Core. While the Up Core has a Cherry Trail Atom x5-Z8350, the UP Core Plus gives you a choice of three Apollo Lake generation quad-core Atom models: the x5-E3930, x5-E3940, and x7-E3950 with TDPs ranging from 6.5W to 12W.

Aaeon has yet to post full specs for the board, which is available with a chassis that offers room for expansion. You can order the board with 2GB (with the Atom x5-E3930), 4GB (E3940), or 8GB (E3950) DDR4-2400, with 32GB eMMC (E3930), 64GB eMMC (E3940 and E3950), or 128GB (E3950). The board is equipped with a DisplayPort for up to 4K, as well as an eDP connection and 2x MIPI-CSI camera interfaces.

The UP Core Plus is further equipped with USB 3.0 host and OTG ports, as well as 2x USB 2.0 host ports. There’s no Ethernet, but you get 802.11ac WiFi. You can also add Ethernet by plugging in the NetPlus board with 4x GbE or the AI Plus FPGA board with one GbE port.

The dual 100-pin high speed connectors that these boards plug into express I/O including I2C, I2S, ISH, UART, PCI-Express, and USB 3.0. The connectors are also compatible with other UP board add-on boards.

The Plus add-on boards include:

AI Plus, back and front
(click image to enlarge)

AI Plus — This FPGA board has the same dimensions as the UP Core Plus. The Cyclone 10GX FPGA has its own 1GB of DDR3, as well as a DisplayPort input for video acquisition and an LVDS input for camera acquisition that is natively compatible to Basler LVDS cameras. The AI Plus is further equipped with single USB 3.0 and GbE ports, as well as a mini-PCIe slot.

The Cyclone 10GX provides between 105K and 220K LEs plus IEEE 754-compliant hard floating-point DSP blocks. The FPGA features a 12.5Gbps chip-to-chip transceiver and 6.6Gbps backplane support. There’s a 1,866Mbps external memory interface and 1.434Gbps LVDS I/O.




Vsion Plus render
(click image to enlarge)

Vision Plus — This vision processing board also appears to have the same dimensions as the UP Core Plus. In addition to its 3x Myriad 2 VPUs, the board provides 512MB RAM, a GbE port, 2x USB 3.0 ports, and a mini-PCIe slot.

Intel’s Movidius Neural Compute Stick

The Myriad 2 2450 VPUs built into the Vision Plus board, as well as the AI Core and AI Core M2, provide neural network acceleration that can translate deep learning frameworks like Caffe and TensorFlow into its own format for rapid prototyping. Available also on the Fathom stick computer and newer Intel Neural Compute Stick, the Linux-oriented Myriad 2 enables the implementation of deep learning technologies directly on embedded, mobile, and IoT devices without requiring access to a cloud-based neural network.

UP AI Edge systems can be configured with up to 5x Myriad 2 chips, says Aaeon. Since it appears you can only combine one Plus add-on board to the SBC at a time, and since none of the boards appear to have an M2 slot to add an AI Core M2 except for the UP Squared, which does not support the Plus add-ons, it would appear this is possible by adding two Intel Neural Compute Sticks to the Vision Board’s USB 3.0 ports.




Net Plus

Net Plus — The Net Plus companion board is equipped with 4x GbE ports. Intel I211AT controllers are standard and I210 controllers with support for TSN are optional. The Net Plus also furnishes a mini-PCIe slot, a USB 3.0 port, and a SATA interface.

 
Further information

UP AI Edge systems are available on Kickstarter through June 30 starting at 159 Euros ($186) for a low-end configuration of the Up Core Plus SBC with chassis and PSU. The combo packages range from 319 Euros ($374) to 679 Euros ($795). The AI Core M2 bundles ship in June, and the UP Squared dev kit ships in August. All the other packages ship in October. More information may be found in Aaeon’s UP AI Edge Kickstarter page.