Menu

Automated Road Network Extraction and Route Travel Time Estimation from Satellite Imagery

SN5: Automated Road Network Extraction and Route Travel Time Estimation from Satellite Imagery

The Problem

Determining optimal routing paths in near real-time is at the heart of many humanitarian, civil, military, and commercial challenges. This statement is as true today as it was two years ago when the SpaceNet Partners announced the SpaceNet Challenge 3 focused on road network detection and routing. In a disaster response scenario, for example, pre-existing foundational maps are often rendered useless due to debris, flooding, or other obstructions. Satellite or aerial imagery often provides the first large-scale data in such scenarios, rendering such imagery attractive.

The SpaceNet 5 challenge sought to build upon the advances from SpaceNet 3 and test challenge participants to automatically extract road networks and routing information from satellite imagery, along with travel time estimates along all roadways, thereby permitting true optimal routing.

The task of this challenge was to output a detailed graph structure with edges corresponding to roadways and nodes corresponding to intersections and end points, with estimates for route travel times on all detected edges. You can find a detailed description of CosmiQ Works’ algorithmic baseline on their blog at The DownLinQ.

SpaceNet open sourced new data sets for the following cities: Moscow, Russia; Mumbai, India; and San Juan, Puerto Rico. For the first time in SpaceNet history, the final submissions were tested on a mystery city dataset that was revealed and open sourced at the end of the Challenge.

Earn AWS GPU Credits!

The first 20 competitors to reach a score of 50 (out of a possible 100) received a credit for 10 hours on a p3.2xlarge for training and improving their models.  To further aid competitors, the SpaceNet 5 baseline is fully open source, and yields a score of 54.  If you have any questions, please reach out through the Topcoder Forum (https://www.topcoder.com/challenges/30099956).

The Data – Over 8,100km of Road Network Labels Across Four New SpaceNet AOIs

AOI Area of Raster (Sq. Km) Road Network Labels (Km)
AOI_7_Moscow 1,353 3,066
AOI_8_Mumbai 1,021 1,951
AOI_9_San Juan 285 1,139
AOI_10_Mystery City ??? ???

Catalog

The data is hosted on AWS as a Public Dataset. It is free to download, but an AWS account is required.
aws s3 ls s3://spacenet-dataset/spacenet/SN5_roads/
In building off of the results from SpaceNet 3, this challenge will use a modified version of the Average Path Length Similarity (APLS) metric that is tuned to optimize travel times between nodes of interest. CosmiQ Works’ blog, The DownLinQ, provides additional information including an APLS metric overview and detailed description of SpaceNet 5 motivation and structure.

Training Data

AOI 7 – Moscow – Road Network Extraction

To download processed ~400mx400m tiles of AOI 7 with associated road network labels for training do the following:

aws s3 cp s3://spacenet-dataset/spacenet/SN5_roads/tarballs/SN5_roads_train_AOI_7_Moscow.tar.gz .

AOI 9 – San Juan – Road Network Extraction

To download processed ~400mx400m tiles of AOI 9 for public testing do the following:

aws s3 cp s3://spacenet-dataset/spacenet/SN5_roads/tarballs/SN5_roads_test_public_AOI_9_San_Juan.tar.gz .

AOI 8 – Mumbai – Road Network Extraction

To download processed ~400mx400m tiles of AOI 8 with associated road network labels for training do the following:

aws s3 cp s3://spacenet-dataset/spacenet/SN5_roads/tarballs/SN5_roads_train_AOI_8_Mumbai.tar.gz .

AOI 10 – Mystery City – Road Network Extraction

To be released after the conclusion of the challenge.

Post-challenge release

Citation Instructions

The SpaceNet Partners, “SpaceNet5: Automated Road Network Extraction and Route Travel Time Estimation from Satellite Imagery”, https://spacenet.ai/sn5-challenge/

License