EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

TO–YOLOX: a pure CNN tiny object detection model for remote sensing images

Zhe Chen, Yuan Liang, Zhengbo Yu, Ke Xu, Qingyun Ji, Xueqi Zhang, Quanping Zhang, Zijia Cui, Ziqiong He, Ruichun Chang, Zhongchang Sun, Keyan Xiao und 1 weitere

International Journal of Digital Earth · 02.10.2023 · 16 (1) · 3882-3904

DOI: 10.1080/17538947.2023.2261901 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Digital Earth, online erschienen am 25.09.2023. Verfasst von Zhe Chen und 12 weiteren Autor:innen. Die Arbeit wurde bisher 25-mal zitiert (Crossref) und nennt 54 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: the Innovative Research Program of the International Research Center of Big Data for Sustainable Development Goals, the Sichuan Natural Resources Department Project, the Project of the Geological Exploration Management Department of the Ministry of Natural Resources.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Remote sensing and deep learning are being widely combined in tasks such as urban planning and disaster prevention. However, due to interference occasioned by density, overlap, and coverage, the tiny object detection in remote sensing images has always been a difficult problem. Therefore, we propose a novel TO–YOLOX(Tiny Object–You Only Look Once) model. TO–YOLOX possesses a MiSo(Multiple-in-Single-out) feature fusion structure, which exhibits a spatial-shift structure, and the model balances positive and negative samples and enhances the information interaction pertaining to the local patch of remote sensing images. TO–YOLOX utilizes an adaptive IOU-T (Intersection Over Uni-Tiny) loss to enhance the localization accuracy of tiny objects, and it applies attention mechanism Group-CBAM (group-convolutional block attention module) to enhance the perception of tiny objects in remote sensing images. To verify the effectiveness and efficiency of TO–YOLOX, we utilized three aerial-photography tiny object detection datasets, namely VisDrone2021, Tiny Person, and DOTA–HBB, and the following mean average precision (mAP) values were recorded, respectively: 45.31% (+10.03%), 28.9% (+9.36%), and 63.02% (+9.62%). With respect to recognizing tiny objects, TO–YOLOX exhibits a stronger ability compared with Faster R-CNN, RetinaNet, YOLOv5, YOLOv6, YOLOv7, and YOLOX, and the proposed model exhibits fast computation.

Förderung

  • the Innovative Research Program of the International Research Center of Big Data for Sustainable Development Goals
  • the Sichuan Natural Resources Department Project
  • the Project of the Geological Exploration Management Department of the Ministry of Natural Resources
  • the Key Research and Development Program of Guangxi
  • the National Natural Science Foundation of China
  • the Chengdu University of Technology Postgraduate Innovative Cultivation Program: Tunnel Geothermal Disaster Susceptibility Evaluation in Sichuan-Tibet Railway…

Zitieren

APA

Chen, Z., Liang, Y., Yu, Z., Xu, K., Ji, Q., Zhang, X., Zhang, Q., Cui, Z., He, Z., Chang, R., Sun, Z., Xiao, K., & Guo, H. (2023). TO–YOLOX: a pure CNN tiny object detection model for remote sensing images. International Journal of Digital Earth, 16(1), 3882-3904. https://doi.org/10.1080/17538947.2023.2261901

Vancouver

Chen Z, Liang Y, Yu Z, Xu K, Ji Q, Zhang X, et al. TO–YOLOX: a pure CNN tiny object detection model for remote sensing images. International Journal of Digital Earth. 2023;16(1):3882-3904. doi:10.1080/17538947.2023.2261901

BibTeX
@article{Chen2023,
  author = {Chen, Zhe and Liang, Yuan and Yu, Zhengbo and Xu, Ke and Ji, Qingyun and Zhang, Xueqi and Zhang, Quanping and Cui, Zijia and He, Ziqiong and Chang, Ruichun and Sun, Zhongchang and Xiao, Keyan and Guo, Huadong},
  title = {TO–YOLOX: a pure CNN tiny object detection model for remote sensing images},
  journal = {International Journal of Digital Earth},
  year = {2023},
  volume = {16},
  number = {1},
  pages = {3882--3904},
  doi = {10.1080/17538947.2023.2261901}
}
Kostenlos starten →