SpaceNet 6: Dataset Release

Preface: SpaceNet LLC is a nonprofit organization dedicated to accelerating open source, artificial intelligence applied research for geospatial applications, specifically foundational mapping (i.e. building footprint & road network detection). SpaceNet is run in collaboration with CosmiQ Works, Maxar Technologies, Intel AI, Amazon Web Services (AWS), Capella Space, Topcoder, and IEEE GRSS.

The SpaceNet partners are proud to announce the release of the SpaceNet 6: Multi-Sensor All Weather Mapping (MSAW) dataset. This openly-licensed dataset features a unique combination of half-meter Synthetic Aperture Radar (SAR) imagery from Capella Space and half-meter electro-optical (EO) imagery from Maxar. We believe this dataset will help to further research and expand remote sensing analytics beyond the EO spectrum and into new modalities. You can download the data now, all you need is an AWS account and the AWS CLI installed and configured. Once you’ve done that, simply run the command below to download the training dataset to your working directory!

aws s3 cp s3://spacenet-dataset/spacenet/SN6_buildings/tarballs/SN6_buildings_AOI_11_Rotterdam_train.tar.gz .

SpaceNet 6 focuses on foundational mapping and building footprint extraction using a combination of these datatypes over the newest and 11th SpaceNet city: Rotterdam, the Netherlands. In this challenge, the training dataset contains both SAR and EO imagery, however the testing and scoring datasets contain only SAR data. Consequently, the EO data can be used for pre-processing the SAR data in some fashion, such as colorization, domain adaptation, or image translation, but cannot be used to directly map buildings. We structure the dataset in such a way to mimic real-world scenarios where historical EO data may be available, but concurrent EO collection with SAR is often not possible due to inconsistent orbits of the sensors, or cloud cover that will render the EO data unusable.

A gif showing a small portion (~5%) of the aerial SAR collects and flight lines over Rotterdam as they happen over time on August 23, 2019. SAR intensity data is visualized with polarizations HH, VV, and HV. Imagery courtesy of SpaceNet partners Capella Space and Maxar Technologies.

The SAR data featured in SpaceNet MSAW is provided by Capella Space via an aerial mounted sensor and closely mimics the spaceborne sensors that will be present on Capella’s future constellation of satellites. The total aerial collect features 204 individual image strips captured from both north and south facing look-angles over a three day span. Each of the image strips features four polarizations (HH, HV, VH, and VV) of data and are preprocessed to display the intensity of backscatter in decibel units at half-meter spatial resolution.

A loop of the four polarizations found in the SpaceNet 6 dataset (left) versus RGB satellite imagery (right). Building footprints are annotated in red outlines, curated via the 3DBAG Dataset. Imagery courtesy of SpaceNet partners Capella Space and Maxar Technologies.

The total dataset contains over 48,000 high-quality building footprint annotations, which are provided via the openly available 3D Basisregistratie Adressen en Gebouwen (3DBAG) dataset [1]. The SpaceNet team performed additional quality control on the labels, removing areas from the dataset that were not properly labeled, and added or removed labels for buildings that were missed or had since been destroyed. Finally we dissolve individual addresses (i.e. apartments and townhomes) that co-exist within a single building to form a single footprint for each structure. The dataset also contains a 3D component from an openly available digital elevation model derived from aerial Light Detection and Ranging (LIDAR). Consequently, for each annotation we report the 75th percentile mean, median, and standard deviation of height in meters. Although the height information will not be used in the challenge, such data can be valuable for future research and analysis on the value of SAR and/or EO to detect the height of objects from an overhead perspective.

Three areas from the SpaceNet 6 MSAW dataset: Left: SAR Intensity (HH, VV, VH). Center: Visible Spectrum Imagery (R,G,B). Right: False Color Composite Imagery (NIR, R, G) Imagery courtesy of SpaceNet partners Capella Space and Maxar Technologies.

Whats Next?

The challenge will kick off on Monday, March 16th and the SpaceNet 6 challenge registration site will go live soon on Topcoder. As always, stay tuned to The DownLinQ and check out https://spacenet.ai for more details on the dataset and challenge.

Remember, you can download the data here:

aws s3 cp s3://spacenet-dataset/spacenet/SN6_buildings/tarballs/SN6_buildings_AOI_11_Rotterdam_train.tar.gz .

Special thanks to Daniel Hogan, Adam Van Etten, Nick Weir, the Capella Space, and Maxar teams for their assistance in preparing the dataset and designing the challenge.

References:
[1] Dukai, B. (Balázs) (2018) 3D Registration of Buildings and Addresses (BAG). 4TU.Centre for Research Data. Dataset. https://doi.org/10.4121/uuid:f1f9759d-024a-492a-b821-07014dd6131c