Robots really do share more human-like qualities than you may think! Good and bad.
Like us humans, a robot’s downfall can come about through being overconfident. That’s the catalyst behind a world-first competition, the Robotic Vision Challenge, launched by the Australian Centre for Robotic Vision.
The competition throws down the gauntlet to the world’s robotics and computer vision research communities to join the Centre’s mission to develop new robotic vision technologies to expand the capabilities of ‘truly useful’ robots.
It challenges competitors to detect objects in video data from high-fidelity simulation of three different types of domestic service robots. Think, a future Rosie from The Jetsons.
Nobody wants an overconfident Rosie…
“Big global competitions have been very effective in computer vision research, but they haven’t really pushed the envelope for the sorts of problems an actual robot encounters in the real world,” said Centre Director Peter Corke.
“This new competition aims to solve that. It’s the world’s first robotic vision challenge; not a computer vision challenge.”
Centre Chief Investigator Niko Sünderhauf, one of the masterminds behind the competition, received a $72,000 Google Faculty Research Award to support its development by a team of Centre researchers based at QUT.
The Robotic Vision Challenge has already been accepted as a key workshop at the world’s largest computer vision forum, the Conference on Computer Vision and Pattern Recognition (CVPR), in June. The four-day conference takes place in Long Beach, California (16-20 June, 2019).
“This is very exciting because we involved global research communities at the concept stage of the challenge last year, seeking ideas and insights at major forums like CVPR and the Robotics: Science and Systems conference,” Dr Sünderhauf said.
“We want to continue that global partnership, bringing robotics and computer vision communities together and encouraging new thinking on problem-solving.”
Dr Sünderhauf said the Robotic Vision Challenge was planned as an annual competition and lasting Centre legacy.
“We will be announcing the winner of the first round of the competition at CVPR 2019, with a $5,000 [AUD] cash prize up for grabs,” he said. “There’s still plenty of time to enter.”
How the Robotic Vision Challenge works?
The new competition focuses on probabilistic object detection.
Competitors must detect objects in cluttered indoor settings, like lounge rooms, kitchens, bathrooms and outdoor living areas, with the added challenge of encountering day and night scenes.
The biggest challenge of the competition, however, is its ‘probabilistic’ aspect for object detection. This novel and unique evaluation rewards accurate estimates of spatial and semantic uncertainty using probabilistic bounding boxes.
Probabilistic object detection is important for robots to safely and effectively work in messy and unpredictable real-world environments.
“Today’s best machine learning methods for object detection are often overly confident in their own knowledge and have no good way of expressing when they don’t really recognise what they see.” said Dr Sünderhauf.
“With this challenge we hope to motivate researchers around the world to develop new probabilistic methods that know when they don’t know.”
Challenge co-organiser Feras Dayoub said the primary objective was to help robots avoid being overconfident about where objects are and what they are, potentially resulting in unsafe or undesired behaviour.
“We believe having a sense of spatial and semantic uncertainty will make robots more cautious and safe around humans in many applications and everyday environments,” Dr Dayoub said.
The Centre’s work to create the competition included development of a new evaluation measure, evaluation protocol and dataset to assess how well object detection methods can estimate spatial and label uncertainty.
The dataset consists of over 56,000 images from 18 simulated indoor video sequences, featuring day and night scenes and simulating three different types of domestic service robots.
Keen to enter the Robotic Vision Challenge? Click here to participate in the competition and access more information about the data and submission format.
Visit a dedicated Robotic Vision Challenge website www.roboticvisionchallenge.org and stay up-to-date on all our news across social media: @roboticvisionAU #roboticvisionau / @robVisChallenge
Did you know?
Competitions or ‘challenges’ are deeply rooted in robotics research and for good reason, driving significant breakthroughs in technological advancement. Think RoboCup; Amazon Robotics Challenge (won by a Centre team in 2017); the DARPA Grand Challenge; and Australia’s UAV Challenge, co-organised by the Centre. Even the toughest test in artificial intelligence (AI), the ‘Turing Test’, devised by Alan Turing in 1950 as a test of a machine’s ability to exhibit intelligent behaviour equivalent to (or indistinguishable from) a human, is a form of competition. It was adapted from a Victorian-style competition called the imitation game.
Click here to download copyright free photos for media use.
Shelley Thomas, Communications Specialist
Australian Centre for Robotic Vision
P: +61 7 3138 4265 | M: +61 416 377 444 | E: firstname.lastname@example.org
About The Australian Centre for Robotic VisionThe Australian Centre for Robotic Vision is an ARC Centre of Excellence, funded for $25.6 million over seven years to form the largest collaborative group of its kind generating internationally impactful science and new technologies that will transform important Australian industries and provide solutions to some of the hard challenges facing Australia and the globe. Formed in 2014, the Australian Centre for Robotic Vision is the world’s first research centre specialising in robotic vision. They are a group of researchers on a mission to develop new robotic vision technologies to expand the capabilities of robots. Their work will give robots the ability to see and understand for the sustainable well-being of people and the environments we live in. The Australian Centre for Robotic Vision has assembled an interdisciplinary research team from four leading Australian research universities: QUT, The University of Adelaide (UoA), The Australian National University (ANU), and Monash University as well as CSIRO’s Data61 and overseas universities and research organisations including INRIA Rennes Bretagne, Georgia Institute of Technology, Imperial College London, the Swiss Federal Institute of Technology Zurich, University of Toronto, and the University of Oxford.
Australian Centre for Robotic Vision
2 George Street Brisbane, 4001
+61 7 3138 7549