ETRI-Knowledge Sharing Plaform

KOREAN
논문 검색
Type SCI
Year ~ Keyword

Detail

Conference Paper A Study on Few-shot Object Detection for Warships Based on Data Generation Using Image Outpainting
Cited 0 time in scopus Share share facebook twitter linkedin kakaostory
Authors
SungWon Moon, Jiwon Lee, Jungsoo Lee, Dowon Nam, Wonyoung Yoo
Issue Date
2023-10
Citation
International Conference on Information and Communication Technology Convergence (ICTC) 2023, pp.1855-1857
Publisher
IEEE
Language
English
Type
Conference Paper
DOI
https://dx.doi.org/10.1109/ICTC58733.2023.10393675
Abstract
With the advent of hyperscale AI based on large amounts of data and supercomputing infrastructure to process them, AI has made remarkable progress in many areas. Among them, AI-based image generation technology has recently developed rapidly, and research on the use of AI training data is also active. Augmenting training data with synthetic image generation can be of great help in AI training for defense and medical applications, where data collection is difficult. In particular, in marine environments, where data collection is more difficult than on land, it is difficult to collect data on various weather conditions, so the application of AI-based image generation technology is very effective. In this paper, several images are generated by image outpainting based on objects in a real image, and the generated images are intended to be used as training data for an object detector.
KSP Keywords
Data Collection, Data generation, Generation technology, Marine environment, Medical Applications, Object detection, large amounts of data, object detector, synthetic image generation, training data, weather conditions