Studiomapp has won the fourth place in the international competition “DIUx xView 2018 Detection Challenge”, organised by the US Department of Defense in partnership with the National Geospatial Intelligence Agency to advance key frontiers in computer vision and develop new solutions for national security with a specific focus on disaster response.
The challenge has been run by the Defence Innovation Unit (DIU) – an organisation whose aim is to put in contact the Department of Defence with innovative technology companies and start-ups in order to solve national defense problems.
When disasters such as Hurricane Irma in 2017 swept a trail of destruction and flooding from the Bahamas to Florida, analysts at the National Geospatial-Intelligence Agency (NGA) were tasked each day with analyzing hundreds of high-resolution satellite images. From damaged buildings to impassable roads and flooded bridges, overhead imagery informed rescue and cleanup efforts.
The abundance of available data from satellites at increasingly high spatial and temporal resolutions and the growing global significance of real-world applications enabled by that imagery provide impetus for creating more sophisticated and robust models and algorithms for overhead object detection.
Applying computer vision to overhead imagery has the potential to detect emerging natural disasters, improve response, quantify the direct and indirect impact — and save lives.
The organisers released a new satellite imagery dataset, called xView, which covers more than 1,400 square kilometers of area around the globe. The images have a ground resolution of 30 cm and were provided by DigitalGlobe, an American commercial vendor of space imagery and geospatial content.
The participants had to apply computer vision techniques and machine learning algorithms to overhead images in the context of disaster response, through a process of object identification. Interviews with disaster response and public safety experts informed xView’s 60 diverse object classes – such as damaged buildings, vehicle lots, utility trucks, tents, ships.
The challenge was focused on accelerating innovations and progress in four computer vision domains:
- Reducing minimum resolution for detection
- Improving learning efficiency
- Enabling discovery of more object classes
- Improving detection of fine-grained classes
The public competition, announced in February 2018, launched in March and officially closed in August, attracted more than four thousand submissions from 100 participants from around the world including companies, universities, and individuals. The top performing algorithms were 300 percent more accurate than the government produced baseline.
Studiomapp among the top 5 solvers in the world
When we heard about the xView competition”, said Angela Corbari – Co-Founder and Chief Operating Officer of Studiomapp, ”we immediately decided to challenge our expertise in the field of remote sensing hoping to give a valid contribution to improve the response in case of natural disasters. Studiomapp is specialised in Location Intelligence, and satellite imagery are one of the key data source for finding global information about the environment”.
Leonardo Alberto Dal Zovo, Co-Founder and Chief Technology Officer of Studiomapp commented: “Impressive results have been obtained using Artificial Intelligence applied in the context of computer vision during the last years, but the analysis of satellite images still presents challenging complexities and solutions that works very well in other domains are not performing that well”. The brain behind this success concludes: “We developed a combination of state of the art Deep Learning models capable of bringing very promising overall results. This award adds another important global acknowledgement to our expertise in geo-intelligence, already trusted by different partners and clients”.
Angela Corbari concludes: “Humanitarian Assistance and Disaster Relief (HADR) will benefit a lot from an improved and automatic analysis of scenes, but also other types of business needs can be solved by using this technology. In fact, we are already expanding our technology offer to answer specific problems ranging from Logistics, Mobility, Urban monitoring to Security using other sources of aerial images, like UAV”.
DIU Official press release:
Credits: Digital Globe/xView Dataset