CASIA OpenIR  > 模式识别国家重点实验室  > 先进时空数据分析与学习
AugFPN: Improving Multi-scale Feature Learning for Object Detection
Guo CX(郭超旭)1,2; Bin Fan1; Qian Zhang3; Shiming Xiang1,2; Chunhong Pan1
2020-06
Conference NameIEEE Proceedings of Computer Vision and Pattern Recognition
Pages12595-12604
Conference Date2020-06-14
Conference Placeonline meeting
Abstract

Current state-of-the-art detectors typically exploit feature pyramid to detect objects at different scales. Among them, FPN is one of the representative works that build a feature pyramid by multi-scale features summation. However, the design defects behind prevent the multi-scale features from being fully exploited. In this paper, we begin by first analyzing the design defects of feature pyramid in FPN, and then introduce a new feature pyramid architecture named AugFPN to address these problems. Specifically, AugFPN consists of three components: Consistent Supervision, Residual Feature Augmentation, and Soft RoI Selection. AugFPN narrows the semantic gaps between features of different scales before feature fusion through Consistent Supervision. In feature fusion, ratio-invariant context information is extracted by Residual Feature Augmentation to reduce the information loss of feature map at the highest pyramid level. Finally, Soft RoI Selection is employed to learn a better RoI feature adaptively after feature fusion. By replacing FPN with AugFPN in Faster RCNN, our models achieve 2.3 and 1.6 points higher Average Precision (AP) when using ResNet50 and MobileNet-v2 as backbone respectively. Furthermore, AugFPN improves RetinaNet by 1.6 points AP and FCOS by 0.9 points AP when using ResNet50 as backbone.

KeywordAugFPN, Object Detection
MOST Discipline Catalogue工学
Indexed ByEI
Funding ProjectBeijing Natural Science Foundation[4162064] ; Young Elite Scientists Sponsorship Program by CAST[2018QNRC001] ; National Natural Science Foundation of China[61773377] ; National Natural Science Foundation of China[91646207] ; National Science Foundation of China[61573352,61876180] ; Major Project for New Generation of AI[2018AAA0100400]
Language英语
Document Type会议论文
Identifierhttp://ir.ia.ac.cn/handle/173211/39183
Collection模式识别国家重点实验室_先进时空数据分析与学习
Corresponding AuthorBin Fan
Affiliation1.Institute of Automation, Chinese Academy of Science
2.School of Artifical Intelligence, University of Chinese Academy of Science
3.Horizon Robotics
Recommended Citation
GB/T 7714
Guo CX,Bin Fan,Qian Zhang,et al. AugFPN: Improving Multi-scale Feature Learning for Object Detection[C],2020:12595-12604.
Files in This Item: Download All
File Name/Size DocType Version Access License
Guo_AugFPN_Improving(631KB)会议论文 开放获取CC BY-NC-SAView Download
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[Guo CX(郭超旭)]'s Articles
[Bin Fan]'s Articles
[Qian Zhang]'s Articles
Baidu academic
Similar articles in Baidu academic
[Guo CX(郭超旭)]'s Articles
[Bin Fan]'s Articles
[Qian Zhang]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[Guo CX(郭超旭)]'s Articles
[Bin Fan]'s Articles
[Qian Zhang]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: Guo_AugFPN_Improving_Multi-Scale_Feature_Learning_for_Object_Detection_CVPR_2020_paper.pdf
Format: Adobe PDF
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.