Loading...
Please wait, while we are loading the content...
Similar Documents
Design-Space Exploration of Biologically-Inspired Visual Object Recognition Algorithms Using CPUs , GPUs , and FPGAs
| Content Provider | Semantic Scholar |
|---|---|
| Author | Sriram, Vinay Tsoi, Kuen Hung Luk, Wayne Cox, David |
| Copyright Year | 2010 |
| Abstract | In recent years, biologically-inspired visual object recognition algorithms – those that aim to mirror the computations performed by the brain's visual system – have emerged as exceptionally promising candidates in object and face recognition research, achieving impressive performance on a range of object and face recognition tasks. While these algorithms typically require a large number of operations per image analyzed, recent advances in many-core parallel computing hardware have enabled practical consideration of relatively large biologically-inspired systems. Here, we focus on a key operation found in this class of models, parallel convolution, and explore the design-space of implementations. Specifically, we compare CPU, GPU and FPGA implementations, spanning SISD, SIMD, and MISD parallelization regimes. We find exceptional speed-ups are possible with GPUs; however, we also find that FPGAs – and MISD configurations in general – remain competitive in a number of domains, particularly when power and space considerations outweigh price. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www2.rowland.harvard.edu/files/rowland/files/2010_sriram_.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
