Paper Info
Title
A software-based dynamic-warp scheduling approach for load-balancing the Viola-Jones face detection algorithm on GPUs
Abstract
Face detection is a key component in applications such as security surveillance and human-computer interaction systems, and real-time recognition is essential in many scenarios. The Viola-Jones algorithm is an attractive means of meeting the real time requirement, and has been widely implemented on custom hardware, FPGAs and GPUs. We demonstrate a GPU implementation that achieves competitive performance, but with low development costs. Our solution treats the irregularity inherent to the algorithm using a novel dynamic warp scheduling approach that eliminates thread divergence. This new scheme also employs a thread pool mechanism, which significantly alleviates the cost of creating, switching, and terminating threads. Compared to static thread scheduling, our dynamic warp scheduling approach reduces the execution time by a factor of 3. To maximize detection throughput, we also run on multiple GPUs, realizing 95.6 FPS on 5 Fermi GPUs.
Year
DOI
Venue
2013
10.1016/j.jpdc.2013.01.012
J. Parallel Distrib. Comput.
Keywords
Field
DocType
thread divergence,multiple gpus,dynamic warp scheduling approach,execution time,detection throughput,thread pool mechanism,fermi gpus,novel dynamic warp scheduling,detection algorithm,viola-jones algorithm,software-based dynamic-warp scheduling approach,static thread scheduling,viola jones,simd
Viola–Jones object detection framework,Thread pool,Fair-share scheduling,Computer science,Load balancing (computing),Scheduling (computing),Parallel computing,Algorithm,SIMD,Thread (computing),Face detection,Distributed computing
Journal
Volume
Issue
ISSN
73
5
0743-7315
Citations 
PageRank 
References 
5
0.42
11
Authors
5
Name
Order
Citations
PageRank
Tan Nguyen1615.22
Daniel Hefenbrock2714.13
Jason Oberg319712.45
Ryan Kastner41779147.73
Scott Baden5382.86