Implementing AI in the production environment
November 05, 2018
Blog
Industrial inference system for the optimization of deep learning training models in the production environment. Schlagw?rter:(immer im Singular!!!)
TANK-870AI – Industrial Inference System |
Artificial intelligence (AI) is also gaining more and more importance in the industrial sector. A common application in the production environment is automatic optical inspection (AOI), which is used for the recognition and control of objects. In order to guarantee a high success rate and reliable defect identification, the database of the underlying training model must be constantly refined. This is realized by means of industrial inference systems. With the new TANK-870AI, ICP Deutschland offers developers such an inference system that convinces with a pre-installed Linux Ubuntu 16.04 LTS operating system and ready-to-use software. The hardware is based on an Intel Skylake or Kaby Lake CPU and has up to 32GB pre-installed RAM. A highlight is the Mustang AI accelerator card in different versions, which is installed via the two PCIe x8 expansion slots. On delivery, the TANK-870AI is already equipped with a 1TB 2.5" HDD and supports another SSD with RAID 0/1 functionality. Ready-to-use software is the Open Visual Inference & Neural Network Optimization Toolkit (OpenVINOTM) from Intel. It combines different tools like the Intel Deep Learning Deployment Toolkit, optimized computer vision libraries, Intel Media SDK OpenCLTM graphics drivers and runtimes as well as current topologies like AlexNet, GoogleNet and more. With OpenVINOTM, deep learning training models such as those from Caffe, MXNET and Tensorflow can be optimized across hardware like CPU, GPU or the Mustang Intel Movidus or FPGA AI acceleration cards. The powerful and scalable inference system TANK-870AI from ICP allows placement of AI application-related and locally at the edge.