Extended Feature Pyramid Network with Adaptive Scale Training Strategy and Anchors for Object Detection in Aerial Images
Multi-scale object detection is a basic challenge in computer vision. Although many advanced methods based on convolutional neural networks have succeeded in natural images, the progress in aerial images has been relatively slow mainly due to the considerably huge scale variations of objects and man...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-03-01
|
Series: | Remote Sensing |
Subjects: | |
Online Access: | https://www.mdpi.com/2072-4292/12/5/784 |
id |
doaj-20d0a07494144627bfaba9206c4dab45 |
---|---|
record_format |
Article |
spelling |
doaj-20d0a07494144627bfaba9206c4dab452020-11-25T02:27:37ZengMDPI AGRemote Sensing2072-42922020-03-0112578410.3390/rs12050784rs12050784Extended Feature Pyramid Network with Adaptive Scale Training Strategy and Anchors for Object Detection in Aerial ImagesWei Guo0Weihong Li1Weiguo Gong2Jinkai Cui3Key Lab of Optoelectronic Technology & Systems of Education Ministry, Chongqing University, Chongqing 400044, ChinaKey Lab of Optoelectronic Technology & Systems of Education Ministry, Chongqing University, Chongqing 400044, ChinaKey Lab of Optoelectronic Technology & Systems of Education Ministry, Chongqing University, Chongqing 400044, ChinaKey Lab of Optoelectronic Technology & Systems of Education Ministry, Chongqing University, Chongqing 400044, ChinaMulti-scale object detection is a basic challenge in computer vision. Although many advanced methods based on convolutional neural networks have succeeded in natural images, the progress in aerial images has been relatively slow mainly due to the considerably huge scale variations of objects and many densely distributed small objects. In this paper, considering that the semantic information of the small objects may be weakened or even disappear in the deeper layers of neural network, we propose a new detection framework called Extended Feature Pyramid Network (EFPN) for strengthening the information extraction ability of the neural network. In the EFPN, we first design the multi-branched dilated bottleneck (MBDB) module in the lateral connections to capture much more semantic information. Then, we further devise an attention pathway for better locating the objects. Finally, an augmented bottom-up pathway is conducted for making shallow layer information easier to spread and further improving performance. Moreover, we present an adaptive scale training strategy to enable the network to better recognize multi-scale objects. Meanwhile, we present a novel clustering method to achieve adaptive anchors and make the neural network better learn data features. Experiments on the public aerial datasets indicate that the presented method obtain state-of-the-art performance.https://www.mdpi.com/2072-4292/12/5/784aerial imagesobject detectionextended feature pyramid network (efpn)adaptive scale training strategyadaptive anchors |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Wei Guo Weihong Li Weiguo Gong Jinkai Cui |
spellingShingle |
Wei Guo Weihong Li Weiguo Gong Jinkai Cui Extended Feature Pyramid Network with Adaptive Scale Training Strategy and Anchors for Object Detection in Aerial Images Remote Sensing aerial images object detection extended feature pyramid network (efpn) adaptive scale training strategy adaptive anchors |
author_facet |
Wei Guo Weihong Li Weiguo Gong Jinkai Cui |
author_sort |
Wei Guo |
title |
Extended Feature Pyramid Network with Adaptive Scale Training Strategy and Anchors for Object Detection in Aerial Images |
title_short |
Extended Feature Pyramid Network with Adaptive Scale Training Strategy and Anchors for Object Detection in Aerial Images |
title_full |
Extended Feature Pyramid Network with Adaptive Scale Training Strategy and Anchors for Object Detection in Aerial Images |
title_fullStr |
Extended Feature Pyramid Network with Adaptive Scale Training Strategy and Anchors for Object Detection in Aerial Images |
title_full_unstemmed |
Extended Feature Pyramid Network with Adaptive Scale Training Strategy and Anchors for Object Detection in Aerial Images |
title_sort |
extended feature pyramid network with adaptive scale training strategy and anchors for object detection in aerial images |
publisher |
MDPI AG |
series |
Remote Sensing |
issn |
2072-4292 |
publishDate |
2020-03-01 |
description |
Multi-scale object detection is a basic challenge in computer vision. Although many advanced methods based on convolutional neural networks have succeeded in natural images, the progress in aerial images has been relatively slow mainly due to the considerably huge scale variations of objects and many densely distributed small objects. In this paper, considering that the semantic information of the small objects may be weakened or even disappear in the deeper layers of neural network, we propose a new detection framework called Extended Feature Pyramid Network (EFPN) for strengthening the information extraction ability of the neural network. In the EFPN, we first design the multi-branched dilated bottleneck (MBDB) module in the lateral connections to capture much more semantic information. Then, we further devise an attention pathway for better locating the objects. Finally, an augmented bottom-up pathway is conducted for making shallow layer information easier to spread and further improving performance. Moreover, we present an adaptive scale training strategy to enable the network to better recognize multi-scale objects. Meanwhile, we present a novel clustering method to achieve adaptive anchors and make the neural network better learn data features. Experiments on the public aerial datasets indicate that the presented method obtain state-of-the-art performance. |
topic |
aerial images object detection extended feature pyramid network (efpn) adaptive scale training strategy adaptive anchors |
url |
https://www.mdpi.com/2072-4292/12/5/784 |
work_keys_str_mv |
AT weiguo extendedfeaturepyramidnetworkwithadaptivescaletrainingstrategyandanchorsforobjectdetectioninaerialimages AT weihongli extendedfeaturepyramidnetworkwithadaptivescaletrainingstrategyandanchorsforobjectdetectioninaerialimages AT weiguogong extendedfeaturepyramidnetworkwithadaptivescaletrainingstrategyandanchorsforobjectdetectioninaerialimages AT jinkaicui extendedfeaturepyramidnetworkwithadaptivescaletrainingstrategyandanchorsforobjectdetectioninaerialimages |
_version_ |
1724841917408083968 |