Title :
A simultaneous multithreading heterogeneous object recognition processor with machine learning based dynamic resource management
Author :
Oh, Jinwook ; Kim, Gyeonghoon ; Park, Junyoung ; Hong, Injoon ; Lee, Seungjin ; Kim, Joo-Young ; Yoo, Hoi-Jun
Author_Institution :
Dept. of Electr. Eng., Korea Adv. Inst. of Sci. & Technol. (KAIST), Daejeon, South Korea
Abstract :
A simultaneous multithreading multicore processor is proposed to accelerate object recognition for 720p HD video streams. The multithreading architecture with Q-learning based dynamic resource management enables concurrent processing of 8 region-of-interests with 5-stage fine grained recognition pipeline outperforming previous object recognition processors with 342GOPS computing power. In addition, the dynamic resource management contributes to increase of energy efficiency by applying the on-line learning DVFS and dynamic tile allocation based on task variance and hardware utilization to achieve 9.6mJ/frame with 1280×720 pixel image. It achieves 2.72× throughput and 3.7× energy efficiency compared to previous recognition processors.
Keywords :
learning (artificial intelligence); multi-threading; object recognition; video signal processing; Q-learning based dynamic resource management; concurrent processing; dynamic tile allocation; energy efficiency; fine grained recognition pipeline; hardware utilization; machine learning; multithreading architecture; online learning DVFS; simultaneous multithreading heterogeneous object recognition processor; simultaneous multithreading multicore processor; task variance; video streams; Computer architecture; Dynamic scheduling; Multithreading; Object recognition; Pipelines; Resource management; Throughput; Simultaneous multithreading; dynamic resource management; multicore processor; object recognition;
Conference_Titel :
Cool Chips XV (COOL Chips), 2012 IEEE
Conference_Location :
Yokohama
Print_ISBN :
978-1-4673-1201-1
Electronic_ISBN :
978-1-4673-1200-4
DOI :
10.1109/COOLChips.2012.6216579
