Focus-and-Detect: A small object detection framework for aerial images


Creative Commons License

Koyun O. C. , Keser R. K. , Akkaya I. B. , Töreyin B. U.

SIGNAL PROCESSING-IMAGE COMMUNICATION, vol.104, 2022 (SCI-Expanded) identifier identifier identifier

  • Publication Type: Article / Article
  • Volume: 104
  • Publication Date: 2022
  • Doi Number: 10.1016/j.image.2022.116675
  • Journal Name: SIGNAL PROCESSING-IMAGE COMMUNICATION
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, Applied Science & Technology Source, Communication Abstracts, Computer & Applied Sciences, INSPEC, Metadex, Civil Engineering Abstracts
  • Keywords: Object detection, Small object detection, Region search, Aerial images
  • Istanbul Technical University Affiliated: Yes

Abstract

Despite recent advances, object detection in aerial images is still a challenging task. Specific problems in aerial images makes the detection problem harder, such as small objects, densely packed objects, objects in different sizes and with different orientations. To address small object detection problem, we propose a two-stage object detection framework called "Focus-and-Detect". The first stage which consists of an object detector network supervised by a Gaussian Mixture Model, generates clusters of objects constituting the focused regions. The second stage, which is also an object detector network, predicts objects within the focal regions. Incomplete Box Suppression (IBS) method is also proposed to overcome the truncation effect of region search approach. Results indicate that the proposed two-stage framework achieves an AP score of 42.06 on VisDrone validation dataset, surpassing all other state-of-the-art small object detection methods reported in the literature, to the best of authors' knowledge.