Dealing with limited remote sensing training data by image synthesis and transfer learning applied to disaster flood mapping

2020 Fiscal Year Annual Research Report

• Project/Area Number: 20K19834
• Research Institution: Institute of Physical and Chemical Research
• Principal Investigator: Baier Gerald 国立研究開発法人理化学研究所, 革新知能統合研究センター, 特別研究員 (80865618)
• Project Period (FY): 2020-04-01 – 2021-03-31
• Keywords: image synthesis / multimodal dataset

Outline of Annual Research Achievements

- Using a generative-adversarial-network (GAN) for remote sensing image synthesis from land cover maps and auxiliary raster information. A submitted journal paper is currently under review https://arxiv.org/abs/2011.11314
- Creation of a multimodal dataset of remote sensing images and making it publicly available https://ieee-dataport.org/open-access/geonrw. The dataset can not only be used for image synthesis but also image segmentation.

Research Products

• [Journal Article] GeoNRW (2020)
  • Author(s): Gerald Baier, Antonin Deschemps, Michael Schmitt, Naoto Yokoya
  • Journal Title: IEEE dataport Volume: online Pages: online
  • DOI: https://dx.doi.org/10.21227/s5xq-b822
  • Open Access / Int'l Joint Research
• [Journal Article] Building a Parallel Universe Image Synthesis from Land Cover Maps and Auxiliary Raster Data (2020)
  • Author(s): Gerald Baier, Antonin Deschemps, Michael Schmitt, Naoto Yokoya
  • Journal Title: IEEE Transactions on Geoscience and Remote Sensing Volume: submitted Pages: submitted
  • Int'l Joint Research
• [Remarks] Created multimodal dataset
  • URL: https://ieee-dataport.org/open-access/geonrw
• [Remarks] arXiv version of TGRS paper
  • URL: https://arxiv.org/abs/2011.11314