Advantech MIC AIx series
The Advantech MIC-AIx series provide a versatile range of fanless box pc's with integrated NVIDIA GPU acceleration. The Scailabel AI manager runs effortlessly on any of the Advantech MIC-AIx devices.
Find the manufacturers device information (and an overview of the various devices in this range) here.
You can easily install the AI manager on any of the Advantech MIC-AIx devices using our simple one-line install.
Please execute the one-line install procedure described here directly on top of the clean Advantech image that is supplied with the MIC. Or, even better, purchase the MIC at one of our resellers with the AI manager pre-installed.
Login to the Edge device and open a shell. Next, type the following command:
sudo bash -c "$(wget -q -O - https://get.sclbl.net)"
This will install the Scailable AI manager on the device. Note that the installation will automatically detect the presence of NVIDIA accelerator hardware and will install all the neccesary drivers. Thus, there is no need to pre-install any of the cuda-drivers or otherwise.
After the installation script has ran you will see the following message:
Before you can use the camera in the AI Manager, available at port :8081 and port :8443, you first need to make it available though the Advantech ICAM UI at port :5000. To do so, go to the ip address suggested by the Scailble one line install script with the port changed to :5000. So for the above example http://192.168.64.2:5000 or http://localhost:5000.
Now, choose "new project", configure the camera to your liking, set it to connect and then play.
Now you can close the browser or browser tab, and continue setting up the AI Manager at port :8081 or :8443. The ICAM version of the AI manager automatically sets its default input to the V4L2 device on /dev/video10 (made available when the camera was set to play in the Advantech ICAM UI).
Depending on which device from the series you select there are virtually no constraints on model size and execution speed on this range of devices.
For AI solutions on the MIC-AIx series we provide the following tips:
Some first-time user fail to appreciate the power of the AI manager running on the device and start coding up (often using some mash-up of docker containers, python glue, and other extremely useful cloud tools) ways of grabbing camera images or resizing images; none of this is neccesary. Once you have connected your camera to the device, you will simply be able to configure the input cameras and off you go.
The MIC AI can have limited filesystem space based on the version of the device. Free up space by removing unneeded large programs from the filesystem with the following commands:
sudo apt update
sudo apt remove thunderbird libreoffice-* -y
sudo apt autoremove -y
sudo apt clean
If you need still more space, try:
sudo rm -rf /usr/local/cuda/samples \
/usr/src/cudnn_samples_* \
/usr/src/tensorrt/data \
/usr/src/tensorrt/samples \
/usr/share/visionworks* ~/VisionWorks-SFM*Samples \
/opt/nvidia/deepstream/deepstream*/samples
Please consult our purchasing options for your region to purchase a desired device with the Scailable AI manager pre-installed. For questions do feel free to reach out to our team anytime using our chat.