SAFDet : a semi-anchor-free detector for effective detection of oriented objects in aerial images
Fang, Zhenyu and Ren, Jinchang and Sun, He and Marshall, Stephen and Han, Junwei and Zhao, Huimin (2020) SAFDet : a semi-anchor-free detector for effective detection of oriented objects in aerial images. Remote Sensing, 12 (19). 3225. ISSN 2072-4292
Preview |
Text (Fang-etal-RS-2020-a-semi-anchor-free-detector-for-effective-detection-of-oriented-objects)
Fang_etal_RS_2020_a_semi_anchor_free_detector_for_effective_detection_of_oriented_objects.pdf Final Published Version License: 
Download (7MB)| Preview |
Abstract
An oriented bounding box (OBB) is preferable over a horizontal bounding box (HBB) in accurate object detection. Most of existing works utilize a two-stage detector for locating the HBB and OBB, respectively, which have suffered from the misaligned horizontal proposals and the interference from complex backgrounds. To tackle these issues, region of interest transformer and attention models were proposed, yet they are extremely computationally intensive. To this end, we propose a semi-anchor-free detector (SAFDet) for object detection in aerial images, where a rotation-anchor-free-branch (RAFB) is used to enhance the foreground features via precisely regressing the OBB. Meanwhile, a center-prediction-module (CPM) is introduced for enhancing object localization and suppressing the background noise. Both RAFB and CPM are deployed during training, avoiding increased computational cost of inference. By evaluating on DOTA and HRSC2016 datasets, the efficacy of our approach has been fully validated for a good balance between the accuracy and computational cost.
ORCID iDs
Fang, Zhenyu, Ren, Jinchang
ORCID: https://orcid.org/0000-0001-6116-3194, Sun, He, Marshall, Stephen ORCID: https://orcid.org/0000-0001-7079-5628, Han, Junwei and Zhao, Huimin;
Item type: Article ID code: 74073 Dates: DateEvent3 October 2020Published29 September 2020AcceptedKeywords: rotate region, convolutional neural network, anchor free, aerial object detection, Electrical engineering. Electronics Nuclear engineering, Electrical and Electronic Engineering Subjects: Technology > Electrical engineering. Electronics Nuclear engineering Department: Faculty of Engineering > Electronic and Electrical Engineering
Technology and Innovation Centre > Sensors and Asset Management
Strategic Research Themes > Measurement Science and Enabling TechnologiesDepositing user: Pure Administrator Date deposited: 05 Oct 2020 08:28 Last modified: 15 Apr 2021 00:42 URI: https://strathprints.strath.ac.uk/id/eprint/74073
CORE (COnnecting REpositories)
CORE (COnnecting REpositories)