DevKit for inference at the edge

March 19, 2019 ICP Deutschland GmbH
TANK-870AI – AIoT Developer Kit
TANK-870AI – AIoT Developer Kit

Deep Learning consists of the sub-areas training and inference. In the training phase, a training model is developed, tested and refined to the desired accuracy using a comprehensive dataset of images and videos. While in the inference phase a rapid and reliable deployment of the training model in the target environment is the focus.

For this purpose, compact edge hardware is recommended that features high-performance, flexible expansions and pre-installed ready-to-use software tools. The new inference system TANK-870AI from ICP Deutschland offers software developers such a platform, namely AIoT Developer Kit. The hardware itself is based on a 6th / 7th generation Intel® Core i5 / i7 or Xeon® E3 CPU with maximum 32GB pre-installed RAM and 1TB SSD mass storage. In addition to numerous common interfaces, two PCIe x8 slots are available to extend functionality and boost the system via FPGA and VPU based AI accelerators.

On the software side, the TANK-870AI is packed with the Open Source Open Visual Inference Neural Network Optimization (OpenVINO™) toolkit. OpenVINO™ enables CNN (convolutional neural network) based, pre-trained models to be used at the edge. The integrated C++ Inference Engine API and the framework friendly Model Optimizer take a big part here. The latter supports multiple frameworks like Caffe, Tensorflow, MXNet and ONNX. Consequently, a wide range of compatible CNN topologies such as AlexNET, SqueezeNet, etc. can be used. In addition, several pre-trained demo sets for common machine vision applications can be used, without the need of time-consuming search and training, for inference in the target environment.

The Intel® Media SDK provides optimized and accelerated decoding, processing and encoding of new media and video data before and after passing through the Inference Engine. The All-in-One Intel® System Studio Suite, though, improves the performance, efficiency and reliability of IoT applications. Another integrated tool is the cloud-based web editor Arduino® Create. It is a plug-in for Intel® platforms with an Ubuntu 16.04 operating system, which is used for project development, collaboration and communication from the idea to the actual application.

On the AIoT Developer Kit from ICP hardware and software are optimally geared to each other. This gives software developers the opportunity to exploit their full deep learning development potential at the edge.

ICP. Industrial Computer Products …by people who care!

Previous Article
Guardian Optical Technologies Demonstrates In-Cabin Sensor Technology Running on NVIDIA DRIVE Platform at GTC

Sensor Solution Offers Automakers Ability to Create Safer Experience for Drivers and Passengers

Next Article
Deep learning inference at the edge
Deep learning inference at the edge

VPU accelerator in PCIe expansion card format for acceleratiing deep learning inference at the edge.