DeePaste - Inpainting For Pasting

Levi Kassel*, Michael Werman

*Corresponding author for this work

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

Abstract

One of the challenges of supervised learning training is the need to procure substantial amount of tagged data. A well-known method of solving this problem is to use synthetic data in a copy-paste fashion so that we cut objects and paste them onto relevant backgrounds. Pasting the objects naively results in artifacts that cause models to give poor results on real data. We present a new method for pasting objects on different backgrounds so that the dataset created gives competitive performance on real data by only treating the border of the pasted object using inpainting. We show state-of-the-art results both on instance detection and foreground segmentation.

Original languageAmerican English
Title of host publicationProceedings - 16th International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2022
EditorsKokou Yetongnon, Albert Dipanda, Luigi Gallo
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages427-433
Number of pages7
ISBN (Electronic)9781665464956
DOIs
StatePublished - 2022
Event16th IEEE International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2022 - Dijon, France
Duration: 19 Oct 202221 Oct 2022

Publication series

NameProceedings - 16th International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2022

Conference

Conference16th IEEE International Conference on Signal-Image Technology and Internet-Based Systems, SITIS 2022
Country/TerritoryFrance
CityDijon
Period19/10/2221/10/22

Bibliographical note

Publisher Copyright:
© 2022 IEEE.

Keywords

  • Copy-paste
  • Inpainting
  • Synthetic data

Fingerprint

Dive into the research topics of 'DeePaste - Inpainting For Pasting'. Together they form a unique fingerprint.

Cite this