Deep learning-based object detection in augmented reality: A systematic review

• Published in: Computers in industry Vol. 139; p. 103661
• Main Authors: Ghasemi, Yalda, Jeong, Heejin, Choi, Sung Ho, Park, Kyeong-Beom, Lee, Jae Yeol
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.08.2022
• Subjects: Augmented reality; Computation platform; Deep learning; Mixed reality; Object detection; Systematic review; Deep learning; Systematic review; Computation platform; Mixed reality; Augmented reality; Object detection
• ISSN: 0166-3615
• DOI: 10.1016/j.compind.2022.103661

Recent advances in augmented reality (AR) and artificial intelligence have caused these technologies to pioneer innovation and alteration in any field and industry. The fast-paced developments in computer vision (CV) and augmented reality facilitated analyzing and understanding the surrounding environments. This paper systematically reviews and presents studies that integrated augmented/mixed reality and deep learning for object detection over the past decade. Five sources including Scopus, Web of Science, IEEE Xplore, ScienceDirect, and ACM were used to collect data. Finally, a total of sixty-nine papers were analyzed from two perspectives: (1) application analysis of deep learning-based object detection in the context of augmented reality and (2) analyzing the use of servers or local AR devices to perform the object detection computations to understand the relation between object detection algorithms and AR technology. Furthermore, the advantages of using deep learning-based object detection to solve the AR problems and limitations hindering the ultimate use of this technology are critically discussed. Our findings affirm the promising future of integrating AR and CV. •Deep learning research in augmented reality over the past decade is reviewed.•The standard object detection algorithms and the applications are summarized.•The computation platforms to implement algorithms in AR are compared.•This review demonstrates the lack of evaluation with human subjects.•Server-side computation is more common when using AR devices.