IIoT scenarios that involve the likes of image processing, text analysis, cyber security require intense computational power. In such cases, Neural Networks are being used, and here’s where Edge Computing & analytics have become essential.

Recently, we worked on a couple of projects where we were required to solve scenarios using Edge Computing devices that can run our AI. We explored available options for edge devices, and based on our conclusions, we’re compared 5 major competitors in the current market.

Firstly though, for context, here’s what we were expected to deploy edge devices for those two projects:

1) In an electroplating plant, to create a Neural Network model for Process Modeling and Control to determine the best control settings for the plant
2) In an OEM manufacturing plant, to automatically inspect defects for Quality Control. Cameras were used in this instance

Like I said, we pored over the pros & cons of multiple options available in the market, and thought we’d chronicle our findings in the table below. We hope you find it useful!

Here we go:

Features/Devices Google Coral(Dev Board) Intel Movidius
(Compute stick 2)
Intel’s UP Squared AI Vision X Developer Kit NVIDIA Jetson (Nano) Raspberry Pi (3 b)
CPU Quad Arm Cortex-A53,Cortex-M4F QuadCore 1.6GHz Atom x7-E3950 QuadCore 1.43GHz ARM A57 QuadCore 1.2GHz Broadcom BCM2837
GPU Integrated GC7000 Lite Graphics Integrated Intel HD Graphics 505 128-core Maxwell GPU N/A
Dedicated HW for Deep Learning, ANN, CV Google Edge TPU coprocessor 16 Core Myriad X (Opt)16 Core Myriad X N/A N/A
RAM 1 GB N/A 8 GB 4 GB 1 GB
Storage 8 GB N/A 64 GB 16 GB Upto 32 GB SD Card
Wireless WiFi 2.4 Ghz, BLE 4.1 N/A N/A N/A WiFi 2.4 Ghz, BLE 4.1
Ethernet 1 N/A 2 1 1
USB 2 Type C, 1 Type A, 1 Micro (3.0) USB 3.0 2 Type A, 1 Micro OTG, 1 Type A 1 USB 3.0, 3 USB 2.0 4 USB 2.0
GPIO 40 N/A 40 40 40
Audio 3.5mm, 4 Pin Stereo N/A N/A N/A 3.5mm
Power 5V USB Type C 5V USB 3.0 5V Power Jack 5V Jack/ PoE/Micro USB 5V Micro USB
Camera Interface MIPI N/A MIPI MIPI
PCIe N/A N/A Yes Yes N/A

These options were analyzed not only for features, but also for compatibility with manufacturing assets, supported OS and so on.

We decided to zero-in on NVIDIA Jetson for those particular implementations, mainly because the Nano is built on a 128 Core Maxwell GPU which is used on GeForce graphic cards. It also supports a lot of ML frameworks with a higher level of performance compared to others.

We rejected the others because:

Google Coral’s TPUs are built for high volume, low precision processing and can only work with Tensorflow.
Intel Up2 with Compute Stick 2 or the compute Stick 2 with any Ubuntu / Raspberry Pi – provides SDK for Computer Vision(Open Vino) which is not supported in windows.

Google Coral is the latest in market and sounds promising. We’ll certainly be keeping an eye on it, and might even recommend it to our clients if there’s a feature fit.

What do you think of the current options? Would you have chosen differently? Let us know if that is the case; we’d love to talk about it.


Have something to add to the conversation? We’re all ears!

Leave a Reply