In-vehicle device for neural networks inference
We work with transport and use neural networks inference at in-vehicle devices. Let’s talk about problems in edge computing of this market. First of all it is operating temperature -40+85. There’re a lot of industrial devices with operating temperature -20+85, but it is a big problem to find insustrial device for Russian winter. The second problem is cost effective solution with high performance. For inference it’s able to use Intel MyriadX VPU as accelerator for edge computing, but all M.2 or other extension boards have 0+ or -20+ operating temperature. For example AAEON AI Core X or VEGA-330.
We decided to produce own inference device with Upsqaured and Movidius inside with climatic module for heating in work and before device can start. This device intended for use in edge computing and has x86 CPU Intel Atom X5 and accelerator for neural networks inference on board.
Let’s look inside the device:
The device has:
- 3G/LTE/GPS module for remote access;
- Intel Movidius (MyriadX) VPU for inference;
- Intel Atom X5 CPU with Intel HD500 iGPU;
- 2 Ethernet RJ45 ports for IP cameras;
- Wide range power supply 12–36V;
- Climatic module.
Small case with passive cooling can give us up to 60 FPS on iGPU in CNN inference and more 45 FPS on Movidius.
This small device can be used as:
- passangers counter in public transport;
- visitors counter in retail;
- poses detector and classifier in industry;
- smokers and cellphone detector for carsharing services;
- driver face recognition for taxi or carsharing services;
- and other cases.
We use Intel OpenVINO toolkit for neural networks inference. It is able to use CPU, iGPU, VPU or other Intel devices for inference in one framework. You can teach your model with GPU servers and convert it in OpenVINO optimizer to use in edge computing.
Read more — https://combox.io/en/projects/outdoor-box-micro/#squared
