View More
View Less
System Message
An unknown error has occurred and your request could not be completed. Please contact support.
Wait Listed
Personal Calendar
Conference Event
There aren't any available sessions at this time.
Conflict Found
This session is already scheduled at another time. Would you like to...
Please enter a maximum of {0} characters.
{0} remaining of {1} character maximum.
Please enter a maximum of {0} words.
{0} remaining of {1} word maximum.
must be 50 characters or less.
must be 40 characters or less.
Session Summary
We were unable to load the map image.
This has not yet been assigned to a map.
Search Catalog
Replies ()
New Post
Microblog Thread
Post Reply
Your session timed out.
This web page is not optimized for viewing on a mobile device. Visit this site in a desktop browser to access the full set of features.
2018 GTC Washington DC

DC8245 - Advancing Computer Vision Frontiers for Disaster Response: The xView Detection Challenge

Session Speakers
Session Description

Earlier this year, the Defense Innovation Unit (DIU) partnered with the National Geospatial-Intelligence Agency (NGA) to introduce xView (, one of the largest publicly available datasets of hand-annotated overhead imagery, and called on experts worldwide to advance key computer vision frontiers for humanitarian assistance and disaster relief. This session will announce the results of the 2018 DIU xView Detection Challenge and share our vision for xView 2.0.

xView 1.0 contains over 1 million labeled objects across 60 diverse classes in imagery captured by DigitalGlobe's WorldView-3 satellites. These classes were chosen for their relevance to humanitarian assistance and disaster relief and included damaged buildings, vehicle lots, utility trucks, tanker trailers, and excavators, marked by bounding boxes. Their selection was based on NGA's work after events such as hurricane Irma, which swept a trail of destruction and flooding from the Bahamas to Florida in 2017.

For the AI community, the xView dataset and challenge motivated progress against four computer vision frontiers: Improving minimum resolution for detection and multi-scale recognition; improving learning efficiency; pushing the limit of discoverable object classes; and improving the detection of fine grained classes. Our challenge resulted in over 2,000 submissions of "live" models (containerized object detection models) being evaluated and scored, representing over 600,000 compute-hours of inference time on the DIU challenge cluster.

The abundance of overhead image data from satellites and the growing diversity and significance of real-world applications enabled by that imagery provide impetus for creating more sophisticated and robust models and algorithms for object detection. We hope xView will become a central resource for a broad range of research in computer vision and overhead object detection, and congratulate our winners!

Additional Information
Computer Vision/Machine Vision
Computer Vision/Machine Vision Deep Learning and AI
Defense General Government / National Labs Higher Education / Research
All technical
50 minutes
Session Schedule