We are creating some awesome events for you. Kindly bear with us.

Machine learning algorithm to automate post-disaster assessments

A computer vision algorithm has been developed by a Defence computer vision specialist. This algorithm has the potential to automate post-disaster assessments and accelerate search and rescue efforts, which will ultimately save lives.

According to a recent press release, the specialist teamed up with researchers at the Australian Institute of Machine Learning (AIML) on a challenge that encourages scientists and mathematicians to create algorithms that will automatically identify objects from satellite images.

Algorithms for satellite imagery

This was a challenging problem because it involved scaling computer vision and machine learning algorithms, which can struggle to cope with the volumes of data found in high-resolution satellite imagery.

Participants were able to submit and test their algorithms against a dataset containing overhead imagery that covers 1,415 square kilometres of complex scenes from around the world.

Even though the specialist has never worked on satellite imagery prior to this, his algorithm ended up ranking second out of more than 4,000 submissions made by 100 participants from around the world.

They discovered that his and other top performing algorithms were 300% more accurate than the existing government algorithms, and are now being incorporated into disaster response systems.

Challenge for large scale computer vision

The team’s algorithm uses deep learning convolutional neural network models designed for much smaller images.

The team members learned important practical lessons for tackling this difficult problem on the journey to their solution.

The algorithm is now being further developed for, among many other things, automated processing of full motion video in support of other R&D programs.

According to Dr Victor Stamatescu, he enjoys tackling difficult intelligence, surveillance and reconnaissance (ISR) problems.

However, he enjoyed the Challenge because it was different to standard research where inputs and the environment can be constrained and controlled and one knows exactly what to expect.

The solutions they created to benefit from leveraging state of the art in computer vision and machine learning techniques.

Defence Science Technology (DST is an excellent environment for this type of research, both within the team and in collaboration with universities.

The Defence Innovation Unit (DIU) announced earlier this year that it had already begun testing winning algorithms from the Challenge in the wake of Hurricane Florence.

It helped assisting emergency personnel in quickly identifying flooded areas and impassable roads.

Who are involved?

The xView Challenge is an initiative of the US Department of Defence’s Defence Innovation Unit (DIU) and the US National Geospatial-Intelligence Agency (NGA).

The Australian Institute of Machine Learning is based at the University of Adelaide. It was officially established in early 2018 through a co-investment by the State Government and the University.

Send this to a friend