A Hybrid FPGA/Coarse Parallel Processing Architecture for Multi-modal Visual Feature Descriptors

Jensen, L.B.W.; Kjær-Nielsen, Anders; Alonso, J.D.; Ros, Eduardo; Krüger, Norbert

doi:10.1109/reconfig.2008.23

Cited by 4 publications

(1 citation statement)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A significant amount of research is engaged with comparisons of the architectures FPGAs, GPUs, and CPUs, and the determination of what problem domains of parallel processing that suits which architecture better (see, e.g., [9]). Previously, some of the processing stages of the system described in this paper have been realized on FPGAs [23,25], and we recognize that there are problem domains in the field of parallel processing, where the FPGA would be the better choice for performance. However, the actual development costs connected to the complexity of FPGA programming as well as the high price of a sufficiently powerful FPGA in combination with the reduced flexibility in respect to code changes connected to required adaptations made FPGAs a less favorable solution in our case.…”

Section: Introductionmentioning

confidence: 99%

A two-level real-time vision machine combining coarse- and fine-grained parallelism

Jensen

Kjær-Nielsen

Pauwels

et al. 2010

J Real-Time Image Proc

Self Cite

View full text Add to dashboard Cite

In this paper, we describe a real-time vision machine having a stereo camera as input generating visual information on two different levels of abstraction. The system provides visual low-level and mid-level information in terms of dense stereo and optical flow, egomotion, indicating areas with independently moving objects as well as a condensed geometric description of the scene. The system operates at more than 20 Hz using a hybrid architecture consisting of one dual-GPU card and one quad-core CPU. The different processing stages of visual information have rather different characteristics that in some cases make fine-grained parallelization on a GPU less applicable. However, for most of the stages that are not efficiently implementable on a GPU, a coarse parallelization on multiple CPU-cores is applicable. We show that with such hybrid parallelism, we can achieve a speed up of approximately a factor 90 and a reduction of latency of a factor 26 compared to processing on a single CPU-core. Since the vision machine provides generic visual information it can be used in many contexts. Currently it is used in a driver assistance context as well as in two robotic applications.

show abstract

Section: Introductionmentioning

confidence: 99%