Online Reinforcement Learning NoC for portable HD object recognition processor

Author

Park, Junyoung ; Hong, Injoon ; Kim, Gyeonghoon ; Oh, Jinwook ; Lee, Seungjin ; Yoo, Hoi-Jun

Author_Institution

Dept. of Electr. Eng., Korea Adv. Inst. of Sci. & Technol. (KAIST), Daejeon, South Korea

fYear

2012

fDate

9-12 Sept. 2012

Firstpage

1

Lastpage

4

Abstract

Heterogeneous multi-core object recognition processor with Reinforcement Learning (RL) NoC is proposed for efficient portable HD object recognition. RL NoC automatically learns management policies in the network of heterogeneous system without an explicit modeling. By adopting RL NoC, the throughput performances of feature detection and description are increased by 20.4% and 11.5%, respectively. As a result, the overall execution time of the object recognition is reduced by 38%. The implemented chip achieves 121mW power consumption with 1.24 TOPS/W power efficiency.

Keywords

feature extraction; learning (artificial intelligence); microprocessor chips; network-on-chip; object recognition; RL NoC; feature description; feature detection; heterogeneous multicore object recognition processor; network-on-chip; online reinforcement learning; portable HD object recognition processor; power 1.24 TW; power 121 mW; Bandwidth; Feature extraction; High definition video; Multicore processing; Object recognition; Resource management; Tiles;

fLanguage

English

Publisher

ieee

Conference_Titel

Custom Integrated Circuits Conference (CICC), 2012 IEEE

Conference_Location

San Jose, CA

ISSN

0886-5930

Print_ISBN

978-1-4673-1555-5

Electronic_ISBN

0886-5930

Type

conf

DOI

10.1109/CICC.2012.6330637

Filename

6330637