Home // INNOV 2023, The Twelfth International Conference on Communications, Computation, Networks and Technologies // View article

Two-Stage Object Detectors: A Comparative Evaluation

Authors:
Jihad Qaddour

Keywords: deep learning; object detection; computer vision; two-stage detectors; and performance analysis.

Abstract:
Object detection is a fundamental task in computer vision with many applications, such as self-driving cars, security, and medical imaging. Recent advances in deep learning have led to significant improvements in the performance of object detectors. This paper presents a comparative performance analysis of generic object detectors, focusing on two-stage detectors. Two-stage detectors are a type of object detector that first generates region proposals and then classifies and refines those proposals. The paper first provides an overview of the taxonomy of two-stage object detection algorithms. It then presents a detailed performance comparison of two-stage detectors on two datasets, Microsoft COCO and PASCAL VOC 2012. The results show that DetectoRS is a state-of-the-art two-stage object detector, outperforming all other two-stage models. However, it is also more complex. The more practical of the two-stage object detectors that performed well in the comparison are Neural Architecture Search-Feature Pyramid Network (NAS-FPN), cascade R-CNN, and Mask R-CNN.

Pages: 1 to 7

Copyright: Copyright (c) IARIA, 2023

Publication date: November 13, 2023

Published in: conference

ISSN: 2326-9286

ISBN: 978-1-68558-104-6

Location: Valencia, Spain

Dates: from November 13, 2023 to November 17, 2023