Parallelization of SIFT on Rigel

Scale-Invariant Feature Transform (SIFT);Rigel;Parallelization;Image recognition

Hussain, Ali A. ; Patel ; Sanjay J.

关键词: Scale-Invariant Feature Transform (SIFT);    Rigel;    Parallelization;    Image recognition;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/24154/Hussain_Ali.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

With the trend towards parallel processing in computing, interest is developing in enabling workloads to be done at faster speeds to enable new usage models. SIFT is an algorithm for image detection and can be used for a variety of purposes. It collects key-point features that are invariant to changes in lighting, orientation and affine transforms. We ported the SIFT algorithm tothe many-core architecture Rigel and studied the amount of speedup that canbe gained by parallelizing the algorithm. Our results showed the algorithm to provide a speedup of 75x when parallelized over 128 cores.