PuSH - Publication Server of Helmholtz Zentrum München

Rekavandi, A.M.* ; Rashidi, S.* ; Boussaïd, F.* ; Hoefs, S.* ; Akbas, E. ; Bennamoun, M.*

Transformers in small object detection: A benchmark and survey of state-of-the-art.

ACM Comput. Surv. 58:64 (2025)
Publ. Version/Full Text DOI
Open Access Hybrid
Creative Commons Lizenzvertrag
Transformers have rapidly gained popularity in computer vision, especially in the field of object detection.Upon examining the outcomes of state-of-the-art object detection methods, we noticed that transformersconsistently outperformed well-established CNN-based detectors in almost every video or image dataset. Smallobjects have been identified as one of the most challenging object types in detection frameworks due to theirlow visibility. This article aims to explore the performance benefits offered by such extensive networks andidentify potential reasons for their Small Object Detection (SOD) superiority. We aim to investigate potentialstrategies that could further enhance transformers’ performance in SOD. This survey presents a taxonomy ofover 60 research studies on developed transformers for the task of SOD, spanning the years 2020 to 2023. Thesestudies encompass a variety of detection applications, including SOD in generic images, aerial images, medicalimages, active millimeter images, underwater images, and videos. We also compile and present a list of 12large-scale datasets suitable for SOD that were overlooked in previous studies and compare the performanceof the reviewed studies using popular metrics such as mean Average Precision (mAP), Frames Per Second(FPS), and number of parameters. Researchers can keep track of newer studies on our web page, which isavailable at: https://github.com/arekavandi/Transformer-SOD.
Altmetric
Additional Metrics?
[➜Log in]
Edit extra informations Login
Publication type Article: Journal article
Document type Review
Keywords Image Stitching ; Popularity
ISSN (print) / ISBN 0360-0300
e-ISSN 1557-7341
Journal ACM Computing Surveys
Quellenangaben Volume: 58, Issue: 3, Pages: , Article Number: 64 Supplement: ,
Publisher Association for Computing Machinery
Reviewing status Peer reviewed
Institute(s) Institute of Environmental Medicine (IEM)