arrow
Return

Learning Orientation-Aware Distances for Oriented Object Detection

delete2023-01-01
delete16
PRE
AI
C
Chaofan Rao
J
Jiabao Wang
G
Gong Cheng *
X
Xingxing Xie
J
Junwei Han
DOI:10.1109/TGRS.2023.3278933delete
deleteOriginal
deleteOriginal request for help
deleteShare
deleteSave
Abstract

Abstract

En 中文
Oriented object detectors have suffered severely from the discontinuous boundary problem for a long time. In this work, we ingeniously avoid this problem by relating regression outputs to regression target orientations. The core idea of our method is to build a contour function which imports orientations and outputs the corresponding distance predictions. Inspired by Fourier transformations, we assume this function can be represented as a linear combination of trigonometric functions and Fourier series. We replace the final 4-D layer in the regression branch of fully convolutional one-stage object detector (FCOS) with a Fourier series transformation (FST) module and term this new network FCOSF. By this unique design, the regression outputs in FCOSF can adaptively vary according to the regression target orientations. Thus, the discontinuous boundary has no impact on our FCOSF. More importantly, FCOSF avoids building complicated oriented box representations, which usually cause extra computations and ambiguities. With only flipping augmentation and single-scale training and testing, FCOSF with ResNet-50 achieves 73.64% mean average precision (mAP) on the DOTA-v1.0 dataset with up to 23.6-frames/s speed, surpassing all one-stage oriented object detectors. On the more challenging DOTA-v2.0 dataset, FCOSF also achieves the highest results of 51.75% mAP among one-stage detectors. More experiments on DIOR-R and HRSC2016 are also conducted to verify the robustness of FCOSF. Code and models will be available at https://github.com/DDGRCF/FCOSF.
Keywords:
Detectors
Object detection
Fourier series
Feature extraction
Training
Predictive models
Transformers
Fourier series transformation (FST)
orientation-aware distance
oriented object detection
remote sensing images

Journal

IEEE Transactions on Geoscience and Remote Sensing cover
IEEE Transactions on Geoscience and Remote Sensing
IF:
8.6
Papers:
2.1W
Citations:
10.7W

Organization

N
Northwestern Polytechnical University
Scholars:
4.6W
Papers: 3.7W
Citations: 5.3W
Cited Papers

Cited Papers

Role of Gamma Knife surgery in the management of pineal region tumors
err2007-12-01
err0
errOAAI
errGregory P. Lekovic; L. Fernando Gonzalez; Andrew G. Shetter; Randall W. Porter; Kris A. Smith; David Brachman; Robert F. Spetzler
errShare
errSave
Conductivity Enhancement in Thin Silicon-on-Insulator Layer Embedding Artificial Dislocation Network
err2011-02-01
err0
PREAI
errYasuhiko Ishikawa; Kazuaki Yamauchi; Chihiro Yamamoto; Michiharu Tabe
errShare
errSave
errShare
errSave
Effect of Altitude, Temperature and Soil on Essential Oil Production inThymus fedtschenkoiFlowers in Osko and Surrounding areas in Iran
err2011-01-01
err0
PREAI
errAbbas Delazar; Masomeh Bahmani; Hasan Hekmat Shoar; A. Tabatabaei-Raisi; Solmaz Asnaashari; Lutfun Nahar; Satyajit Dey Sarker
errShare
errSave
Align Deep Features for Oriented Object Detection
err2022-01-01
err492
errOAAI
errHan, Jiaming; Ding, Jian; Li, Jie; Xia, Gui-Song
errShare
errSave
Dual-Aligned Oriented Detector
err2022-01-01
err118
PREAI
errCheng, Gong; Yao, Yanqing; Li, Shengyang; Li, Ke; Xie, Xingxing; Wang, Jiabao; Yao, Xiwen; Han, Junwei
errShare
errSave
Fewer is more: efficient object detection in large aerial images
err2023-12-18
err45
errOAAI
errXie, Xingxing; Cheng, Gong; Li, Qingyang; Miao, Shicheng; Li, Ke; Han, Junwei
errShare
errSave
researcher View more