Unsupervised harmonious image composition for disaster victim detection

Ning Zhang*, F.C. Nex, G. Vosselman, N. Kerle

*Corresponding author for this work

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

3 Citations (Scopus)
104 Downloads (Pure)


Deep detection networks trained with a large amount of annotated data achieve high accuracy in detecting various objects, such as pedestrians, cars, lanes, etc. These models have been deployed and used in many scenarios. A disaster victim detector is very useful when searching for victims who are partially buried by debris caused by earthquake or building collapse. However, considering that larger quantities of real images with buried victims are difficult to obtain for training, a deep detector model cannot give full play to its advantages. In this paper we generate realistic images for training a victim detector. We first randomly cut out human body parts from an open source human data set and paste them into the ruins background images. Then, we propose an unsupervised generative adversarial network (GAN) to harmonize the body parts to fit the style (illumination, texture and color characteristics) of the background. These generated images are finally used to fine-tune a detection network YOLOv5. We evaluate both the AP (average precision) for IoU (Intersection over Union) 0.5 and for IoU ? [0.5:0.05:0.95], which are denoted as APi@0:5 and APi@.5 : .95], respectively. The best experimental results show that the YOLOv5l pre-trained on the COCO data set performs poorly on detecting victims, and the APi@.5 : .95] is only 19.5%. The model that uses our composite images as fine-tuning data can effectively detect victims, and increases the APi@.5 : .95] to 33.6%. The APi@0:5 increases from 32.4% to 53.4%. Our unsupervised harmonization method further improves the results by 2.1% and 6.1%, respectively.

Original languageEnglish
Title of host publicationThe International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Subtitle of host publicationCongress “Imaging today, foreseeing tomorrow”, Commission III
Number of pages8
Publication statusPublished - 31 May 2022
EventXXIV ISPRS Congress 2022: Congress “Imaging today, foreseeing tomorrow” - Nice, France
Duration: 6 Jun 202211 Jun 2022

Publication series

NameThe International Archives of the Photogrammetry, Remote Sensing and Spatial Information
ISSN (Print)1682-1750
ISSN (Electronic)2194-9034


ConferenceXXIV ISPRS Congress 2022
Abbreviated titleISPRS 2022
Internet address




Dive into the research topics of 'Unsupervised harmonious image composition for disaster victim detection'. Together they form a unique fingerprint.

Cite this