Detecting Objects from No-Object Regions: A Context-Based Data Augmentation for Object Detection

Authors

Jun Zhang¹, Feiteng Han^{1, *}, Yutong Chun^{1, *,} , Kangwei Liu², Wang Chen³

¹School of Management and Engineering, Capital University of Economics and Business, No. 121, Zhangjia Road, Huaxiang, Fengtai District, Beijing, 100070, China

²Alibaba DAMO Academy, District 4, Wangjing East Park, Chaoyang District, Beijing, 100102, China

³School of Information, Renmin University of China, No.59 Zhongguancun Road, Haidian District, Beijing, 100872, China

^*Corresponding authors. Email: h_feiteng@163.com; chunyutong@yeah.net

DOI

10.2991/ijcis.d.210622.003How to use a DOI?

Keywords

Data augmentation; Object detection; Convolutional neural network; Deep learning

Abstract

Data augmentation is an important technique to improve the performance of deep learning models in many vision tasks such as object detection. Recently, some works proposed the copy-paste method, which augments training dataset by copying foreground objects and pasting them on background images. By designing a learning-based context model to predict realistic placement regions, these approaches have been proved to be more effective than traditional data augmentation methods. However, the performance of the existing context model was limited by three problems: (1) The definitions of positive and negative samples generate too much label noise. (2) The examples with masked regions lose a lot of context information. (3) The sizes (i.e., scale and aspect ratios) of predicted regions are sampled from a prior shape distribution, which leads to a coarse estimation. In this work, we first explore the placement rules that generate realism and effective training examples for detectors. And then, we propose a trainable context model in order to find proper placement regions by classifying and refining dense prior default boxes. We also design a corresponding reasonable generation for training examples by annotating ground truth on free space according to the placement rules. The experimental results on PASCAL VOC show that our approach outperforms the state-of-the-art- related work.

Copyright

© 2021 The Authors. Published by Atlantis Press B.V.

Open Access

This is an open access article distributed under the CC BY-NC 4.0 license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)
View full text (HTML)

TY  - JOUR
AU  - Jun Zhang
AU  - Feiteng Han
AU  - Yutong Chun
AU  - Kangwei Liu
AU  - Wang Chen
PY  - 2021
DA  - 2021/06/28
TI  - Detecting Objects from No-Object Regions: A Context-Based Data Augmentation for Object Detection
JO  - International Journal of Computational Intelligence Systems
SP  - 1871
EP  - 1879
VL  - 14
IS  - 1
SN  - 1875-6883
UR  - https://doi.org/10.2991/ijcis.d.210622.003
DO  - 10.2991/ijcis.d.210622.003
ID  - Zhang2021
ER  -