Proceedings of the 6th Computer Science Research Days, JRI 2023, 18-20 December 2023, Ouagadougou, Burkina Faso

Research Article

Comparison of Joblib and Pymp for Parallel Fingerprint Recognition

Download121 downloads
  • @INPROCEEDINGS{10.4108/eai.18-12-2023.2348107,
        author={Ali Zerbo and Mo\~{n}se Ouedraogo and Abdoulaye Sere and Mamadou Diarra},
        title={Comparison of Joblib and Pymp for Parallel Fingerprint Recognition},
        proceedings={Proceedings of the 6th Computer Science Research Days, JRI 2023, 18-20 December 2023, Ouagadougou, Burkina Faso},
        publisher={EAI},
        proceedings_a={JRI},
        year={2024},
        month={6},
        keywords={fingerprint recognition parallel computing joblib pymp cpu},
        doi={10.4108/eai.18-12-2023.2348107}
    }
    
  • Ali Zerbo
    Moïse Ouedraogo
    Abdoulaye Sere
    Mamadou Diarra
    Year: 2024
    Comparison of Joblib and Pymp for Parallel Fingerprint Recognition
    JRI
    EAI
    DOI: 10.4108/eai.18-12-2023.2348107
Ali Zerbo1,*, Moïse Ouedraogo1, Abdoulaye Sere1, Mamadou Diarra2
  • 1: Université Nazi BONI
  • 2: Ecole Polytechnique de Ouagadougou
*Contact email: alizerbo98@gmail.com

Abstract

Fingerprint recognition is a cornerstone technology in security and identification systems, valued for its reliability and uniqueness. As the complexity of fingerprint data increases, efficient computational techniques become crucial to ensure fast and accurate processing. Parallel computing emerges as a promising solution, distributing computational tasks across multiple processors to enhance performance and reduce processing times. This study compares two par- allel computing libraries, Joblib and Pymp, to assess their effectiveness in optimizing fingerprint recognition algorithms. Joblib is renowned for its ease of integration, memory efficiency, and caching support, making it suitable for machine learning tasks and data preprocessing. Pymp, on the other hand, offers a straightforward API for parallelizing loops and managing shared resources, ideal for tasks that require shared memory. Implementing fingerprint recognition processes with both libraries, we measured their performance in terms of execution time, resource utilization, and ease of use. Contrary to expectations, our results show that Pymp surpasses Joblib in speed, even with a moderate dataset of 407 fingerprint images, thanks to its efficient CPU resource management and flexible parallel loop execution.