BROWSE
Detail
Conference Paper
Inference Engine for Embedded Systems Supporting Multi-processors
Cited - time in 
Share 
Share
- Authors
- Seung-mok Yoo, Changsik Cho, Kyung Hee Lee, Jaebok Park, Seok Jin Yoon, Youngwoon Lee, Byung-Gyu Kim
- Issue Date
- 2020-12
- Citation
- International Conference on Interdisciplinary Research on Computer Science, Psychology, and Education (ICIPE) 2020, v.4, no.1, pp.157-163
- Language
- English
- Type
- Conference Paper
- Abstract
- In this paper, we introduce the progress of our project on developing an open deep learning inference engine for embedded systems. The engine was designed to be installed in autonomous vehicles to perform fast neural network inference such as image recognition. Due to the nature of the embedded system, various types of hardware can be used. Some AI applications, e.g., object detection in autonomous vehicle, should be handled in real time. To meet these requirements, our inference engine was designed to operate on multiple OpenCL devices. It is because common of the shelf GPUs support OpenCL. This paper also describes briefly how we implemented the multi-processor support part, which was introduced in the previous work. It shows performance improvements as the number of processors for arithmetic computing changes.
- KSP Keywords
- AI Applications, Autonomous vehicle, Image recognition, Inference engine, Multi-processor, Real-time, Support part, deep learning(DL), embedded system, network inference, neural network(NN)