Search

For conventional SfM pipeline, image matching is enduring limitation when considering the time efficiency. In the last few years, to speed up image matching procedure, many image retrieval works were proposed to fast find overlapping image pairs, e.g., bag-of-word that clusters hand-crafted local features in a hierarchical way for efficient similar image retrieval, or learning-based global features (such as, VGG or ResNet) are used to represent image in a global compact manner. However, there are rarely benchmarks with referenced overlapping information to: first, evaluate the retrieval performance; second, fine tune deep-learning models along the direction that is more capable to deal with overlapping image pairs. In this work, based on traditional photogrammetric procedures, relevant photogrammetric information is obtained including image orientation parameters, 3D mesh model and etc., we then generate a benchmark for determining Overlapping Images - BeDOI, in which referenced pairwise overlapping relationships are estimated via rigorous photogrammetric geometry. To extend the generality, in total, BeDOI contains 13667 images which are basically UAV and close-range images of various scene categories, e.g., urban cities, campus, village, historical relics, green land, buildings and etc. Lastly, to demonstrate the efficacy of the proposed BeDOI, several image retrieval methods are tested and the experimental results are reported as a competition challenge1

In the era of deep learning, annotated datasets have become a crucial asset to the remote sensing community. In the last decade, a plethora of different datasets was published, each designed for a specific data type and with a specific task or application in mind. In the jungle of remote sensing datasets, it can be hard to keep track of what is available already. With this paper, we introduce EOD - the IEEE GRSS Earth Observation Database (EOD) - an interactive online platform for cataloguing different types of datasets leveraging remote sensing imagery.

We present here the scientific outcomes of the 2021 Data Fusion Contest (DFC2021) organized by the Image Analysis and Data Fusion Technical Committee of the IEEE Geoscience and Remote Sensing Society. DFC2021 was dedicated to research on geospatial artificial intelligence (AI) for social good with a global objective of modeling the state and changes of artificial and natural environments from multimodal and multitemporal remotely sensed data toward sustainable developments. DFC2021 included two challenge tracks: “Detection of settlements without electricity” and “Multitemporal semantic change detection.” This article mainly focuses on the outcome of the multitemporal semantic change detection track. We describe in this article the DFC2021 dataset that remains available for further evaluation of corresponding approaches and report the results of the best-performing methods during the contest.

The synergistic combination of deep learning (DL) models and Earth observation (EO) promises significant advances to support the Sustainable Development Goals (SDGs). New developments and a plethora of applications are already changing the way humanity will face the challenges of our planet. This article reviews current DL approaches for EO data, along with their applications toward monitoring and achieving the SDGs most impacted by the rapid development of DL in EO. We systematically review case studies to achieve zero hunger, create sustainable cities, deliver tenure security, mitigate and adapt to climate change, and preserve biodiversity. Important societal, economic, and environmental implications are covered. Exciting times are coming when algorithms and Earth data can help in our endeavor to address the climate crisis and support more sustainable development.

This article presents the scientific outcomes of the 2020 Data Fusion Contest (DFC2020) organized by the Image Analysis and Data Fusion Technical Committee of the IEEE Geoscience and Remote Sensing Society. The 2020 Contest addressed the problem of automatic global land-cover mapping with weak supervision, i.e., estimating high-resolution semantic maps while only low-resolution reference data are available during training. Two separate competitions were organized to assess two different scenarios: 1) high-resolution labels are not available at all; and 2) a small amount of high-resolution labels are available additionally to low-resolution reference data. In this article, we describe the DFC2020 dataset that remains available for further evaluation of corresponding approaches and report the results of the best-performing methods during the contest.