Data augmentation using synthesized images for object detection

Hyunjun Jo, Yong Ho Na, Jae-Bok Song

Research output: Chapter in Book/Report/Conference proceedingConference contribution

7 Citations (Scopus)


Recently deep learning-based research has been conducted in various fields. Deep learning algorithms require vast amounts of data for good performance. Therefore, collecting such a huge amount of high-quality data is crucial to the deep learning-based methods. Data collection is simple but very time-consuming. To cope with this difficulty, in this study we propose a method to generate a dataset by synthesizing the images of background and object. Various images can be generated through post-processes such as adding noise and changing brightness to the images of objects obtained from different viewpoints. Furthermore, we do not need to manually annotate the dataset for object detection because we can calculate the parameters of the bounding boxes from the location and size of object images during the synthesis process. Faster R-CNN, one of the deep learning algorithms for object recognition, was used to verify the proposed method. The performance based on the dataset generated by the proposed method is comparable to that based on the real dataset.

Original languageEnglish
Title of host publicationICCAS 2017 - 2017 17th International Conference on Control, Automation and Systems - Proceedings
PublisherIEEE Computer Society
Number of pages4
ISBN (Electronic)9788993215137
Publication statusPublished - 2017 Dec 13
Event17th International Conference on Control, Automation and Systems, ICCAS 2017 - Jeju, Korea, Republic of
Duration: 2017 Oct 182017 Oct 21


Other17th International Conference on Control, Automation and Systems, ICCAS 2017
CountryKorea, Republic of


  • Data augmentation
  • Deep learning
  • Object detection
  • Synthesized images

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Science Applications
  • Control and Systems Engineering
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Data augmentation using synthesized images for object detection'. Together they form a unique fingerprint.

Cite this