수중 이미지 향상을 통한 해양 쓰레기 검출 및 확률지도 작성

Abstract

This paper introduces the system of reconstructing underwater trash map. First, our systems improve performance of underwater object detection by using deep learning-based image enhancement. Also, we prove that using enhanced image, it needs a little labeled data compare to underwater images. Finally, we reconstruct underwater 2D trash map by stitching images and detecting trash coordinates.