Make Scale Invariant Feature Transform “Fly” with CUDA

Authors

  • Yuhong Mo Carnegie Mellon University, Electrical and Computer Engineering, PA, USA
  • Chaoyi Tan Northeastern University, Electrical and Computer Engineering, MA, USA
  • Chenghao Wang Georgia Institute of Technology, Computer Science, GA, USA
  • Hao Qin Independent, CHINA
  • Yushan Dong University of Maryland, Machine Learning, USA

DOI:

https://doi.org/10.5281/zenodo.11516606

Keywords:

SIFT, CUDA, Parallelism

Abstract

This paper introduces an implementation of scale invariant feature transform (SIFT) algorithm with CUDA. Primary  steps  including  building the  Gaussian  pyramid  and the difference of Gaussian pyramid, identification, localization [1], and orientation  generation of key-points  are  realized  on GPU with CUDA. A detailed description of important kernel function implementations is covered along with optimizations made to achieve high performance, and a comparison between the CUDA version SIFT algorithm and a baseline sequential CPU implementation is included.

Downloads

Download data is not yet available.

Published

2024-06-07

How to Cite

Yuhong Mo, Chaoyi Tan, Chenghao Wang, Hao Qin, & Yushan Dong. (2024). Make Scale Invariant Feature Transform “Fly” with CUDA. International Journal of Engineering and Management Research, 14(3), 38–45. https://doi.org/10.5281/zenodo.11516606

Issue

Section

Articles