ICP Deutschland provides the ultra-compact ITG-100AI, an inference system that is prepared for use with dedicated neural network topologies (DNN). The two Intel® Movidius™ Myriad™ X VPU in the ITG-100AI deliver great inference performance per watt, 16 SHAVES cores for AI calculations and native FP16 support. On the software side, the ITG-100AI supports Intel®'s open source toolkit "Open Visual Inference Neural Network Optimization" (OpenVINO™), delivering end-to-end acceleration for a broad range of neural networks. OpenVINO™ enables CNN (convolutional neural network) based, pre-trained models to be deployed at the edge with little effort.
Training models from Caffe or Tensorflow can be implemented quickly and easily. The ITG-100AI is also equipped with an Intel® Atom™ x5-E3930 Apollo Lake SoC and 8GB pre-installed DDR3L memory. A 128GB SATA DOM provides space for the Linux or Windows 10 operating system as well as for the OpenVINOTM Toolkit, which is available free of charge, and the training model to be used. For the connection of peripheral devices, the palm-sized embedded PC has two LAN GbE, RS-232/422/485 and USB 3.0 interfaces each. The ITG-100AI can be operated in a temperature range from -20°C to +50°C. The compact size and the mounting options offer additional flexible application scope to realize AI projects.