Delving into the Scale Variance Problem in Object Detection

Junliang Chen, Xiaodong Zhao, Linlin Shen

Research output: Chapter in Book/Conference proceedingConference contributionpeer-review

Abstract

Object detection has made substantial progress in the last decade, due to the capability of convolution in extracting local context of objects. However, the scales of objects are diverse and current convolution can only process single-scale input. The capability of traditional convolution with a fixed receptive field in dealing with such a scale variance problem, is thus limited. Multiscale feature representation has been proven to be an effective way to mitigate the scale variance problem. Recent researches mainly adopt partial connection with certain scales, or aggregate features from all scales and focus on the global information across the scales. However, the information across spatial and depth dimensions is ignored. Inspired by this, we propose the multi-scale convolution (MSConv) to handle this problem. Taking into consideration scale, spatial and depth information at the same time, MSConv is able to process multi-scale input more comprehensively. MSConv is effective and computationally efficient, with only a small increase of computational cost. For most of the single-stage object detectors, replacing the traditional convolutions with MSConvs in the detection head can bring more than 2.5% improvement in AP (on COCO 2017 dataset), with only 3% increase of FLOPs. MSConv is also flexible and effective for two-stage object detectors. When extended to the mainstream two-stage object detectors, MSConv can bring up to 3.0% improvement in AP. Our best model under single-scale testing achieves 48.9% AP on COCO 2017 test-dev split, which surpasses many state-of-the-art methods.

Original languageEnglish
Title of host publicationProceedings - 2021 IEEE 33rd International Conference on Tools with Artificial Intelligence, ICTAI 2021
PublisherIEEE Computer Society
Pages902-909
Number of pages8
ISBN (Electronic)9781665408981
DOIs
Publication statusPublished - 2021
Externally publishedYes
Event33rd IEEE International Conference on Tools with Artificial Intelligence, ICTAI 2021 - Virtual, Online, United States
Duration: 1 Nov 20213 Nov 2021

Publication series

NameProceedings - International Conference on Tools with Artificial Intelligence, ICTAI
Volume2021-November
ISSN (Print)1082-3409

Conference

Conference33rd IEEE International Conference on Tools with Artificial Intelligence, ICTAI 2021
Country/TerritoryUnited States
CityVirtual, Online
Period1/11/213/11/21

Keywords

  • multi-scale convolution
  • object detection
  • scale variance

ASJC Scopus subject areas

  • Software
  • Artificial Intelligence
  • Computer Science Applications

Fingerprint

Dive into the research topics of 'Delving into the Scale Variance Problem in Object Detection'. Together they form a unique fingerprint.

Cite this