期刊
2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021
卷 -, 期 -, 页码 5887-5896出版社
IEEE COMPUTER SOC
DOI: 10.1109/CVPR46437.2021.00583
关键词
-
资金
- ANR [ANR-17-CE23-0009]
- Inria Sophia Antipolis - Mediterranee Nef computation cluster
- Army Research Office [W911NF2010168]
- Air Force Office of Scientific Research award [FA9550-19-1-031]
- National Science Foundation [IIS-1838071, 1122374]
- CSAIL Systems that Learn program
- MIT-IBM Watson AI Laboratory
- Toyota-CSAIL Joint Research Center
- MIT.nano Immersion Lab/NCSOFT Gaming Program seed grant
- Skoltech-MIT Next Generation Program
- Agence Nationale de la Recherche (ANR) [ANR-17-CE23-0009] Funding Source: Agence Nationale de la Recherche (ANR)
The research team enhanced a deep segmentation model by adding frame field output to facilitate the conversion of raster segmentations to vector polygons required by geographic information systems. Training a neural network to align predicted frame fields with ground truth contours improved segmentation quality and enabled polygonization.
While state of the art image segmentation models typically output segmentations in raster format, applications in geographic information systems often require vector polygons. To help bridge the gap between deep network output and the format used in downstream tasks, we add a frame field output to a deep segmentation model for extracting buildings from remote sensing images. We train a deep neural network that aligns a predicted frame field to ground truth contours. This additional objective improves segmentation quality by leveraging multi-task learning and provides structural information that later facilitates polygonization; we also introduce a polygonization algorithm that that utilizes the frame field along with the raster segmentation.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据