Copy
Robotic Vision Quarterly Newsletter - March 2018
View this email in your browser

creating robots that see

One of our workshops during the 4th Robotic Vision Summer School (RVSS) held at ANU's Kioloa Campus

DIRECTOR'S MESSAGE
Welcome to our first public newsletter for 2018.  Last year was a major one for the Centre with a number of significant achievements, and this year promises to be just as exciting.
 
The Australian Research Council conducted a formal mid-term review of the Centre last September and comments and recommendations were received just before Christmas.  The report was very positive and recommends that funding be “continued at the current level until the end of currently scheduled funding in 2020”.  This is a great outcome for our Centre and I thank and congratulate everybody in the Centre who has contributed to our success. 
 
We now have a little under three years of funding to build on this achievement.  I see our priorities as continuing to deliver on our research program, creating demonstrators that powerfully showcase our research, and engaging with end-users and deliver benefit to Australia.   ​

Last November I convened a meeting of the Centre’s chief investigators to decide on the science priorities for the remainder of the Centre.  The new research projects we will run from the start of 2018, and their leaders, are:
  • Viorela Ila - Fast Vision-Based Navigation in Unstructured Environments
  • Stephen Gould - Robots, Human and Action
  • Ian Reid - Scene Understanding
  • Anton van den Hengel - Vision and Language
  • Juxi Leitner - Manipulation and Vision
  • Gustavo Carneiro - Learning
  • Niko Sunderhauf - Robotic Vision Virtual Evaluation Testbed 
  • Feras Dayoub - Robotic Vision Deployment and Evaluation 
We also created a new class of project: demonstrator projects.  These are projects where we integrate technology from different research projects to create a system that showcases, in an accessible and compelling way, what robotic vision is all about.  The Amazon Robotics Challenge, last year, is a great example of this. An important benefit of investing in demonstrators is the learnings that will come from robotic vision systems operating in the real world.  These will provide invaluable insight into the performance of perception algorithms and this will inform our research in a way that datasets such ImageNet are unable to do.

The new demonstrator projects that we will run from the start of 2018, and their leaders, are:
  • Peter Corke - Manipulation 
  • Belinda Ward - Social Robotics
  • Michael Milford - Self-driving cars
These projects will be managed by a revamped Research Committee (ResComm) comprising all the people named above, who will meet monthly to review project progress and resolve roadblocks.

A key early focus for the Centre this year is recruiting for these new projects.  Other priorities include establishing an effective governance structure for these projects, completing the robotic vision roadmap (
see below) and preparing for our next annual RoboVis conference which will be held in Canberra having completed a 4-year tour of all the nodes.


Last year we decided to merge our End User Advisory Board and Centre Advisory Committee into a new Centre Advisory Board (CAB). I’d like to welcome our new members - Kylie Ahern and Henrik Christensen who join Alex Zelinsky, Trent Lund, Russel Rankin, Andrew Harris, Sue Keay and myself to complete the board. My huge thanks on behalf of the Centre Executive to our exiting members for their advice, support and service during the first half of the Centre’s operation. I look forward to working with our CAB over the second half of the life of the Centre to achieve great scientific and industrial outcomes. You can view our Board member profiles on our website here.

March 8th was International Women’s Day and a timely reminder for us to consider what actions we can take this year to progress gender parity. We launched our 
Gender Equity Plan at RoboVis in 2017 and I’m pleased to see the number of women in the Centre increasing – welcome to Associate Investigator Professor Elizabeth Croft, Centre Advisory Board member Kylie Ahern and our four new PhD researchers Dimity Miller, Serena Mou, Natalie Jablonsky and Shin Fang Ch’ng. It’s important for all of us to think, act and be gender inclusive. The theme for 2018 is Press for Progress and featured below are three colleagues including myself and our pledge for how we will press for progress in 2018.

Finally, I’m delighted to welcome back Chief Investigators Hongdong Li and Steven Gould who return to Centre after sabbatical periods at Carnegie-Mellon University and Amazon respectively. 

 
As always, up to date information about Centre achievements can be found on our website and our YouTube channel. If you have comments about the newsletter, questions about what we do or how to engage with us, please feel free to email me.
 
Enjoy the issue.

Distinguished Professor Peter Corke
Centre Director 
NEW MEMBERS
We are delighted to welcome our new Centre Members: Associate Invesigator Professor Elizabeth Croft; Centre Advisory Board Members Kylie Ahern and Professor Henrik Christensen.
Professor Elizbeth Croft is an internationally recognised expert in the field of human robotic interaction. Her research interests include Assistive Robotics, Intelligent Robotic Systems, Motional Planning, and Mechatronics. She has led large-scale collaborative research projects with major industry partners and has an exceptional track record in advancing women’s representation and participation in engineering.  View full profile 
Kylie Ahern is an award-winning science publisher and entrepreneur with experience that spans media, telecommunications, science and education. In 2004 she co-founded Cosmos Media and launched Australia’s top-selling science magazine, Cosmos. The company was recognised through 54 awards and commendations for the high quality of its publications, websites, science outreach programs, publishing and journalism. Kylie created educational products – teachers notes, study guides, posters, career guides – to which 70% of Australia’s high schools and hundreds of overseas subscribed to reach more than 130,000 students.
View full profile
Dr Henrik I. Christensen is a Professor of Computer Science at Dept. of Computer Science and Engineering UC San Diego. He is also the director of the Institute for Contextual Robotics. Dr Christensen does research on systems integration, human-robot interaction, mapping and robot vision. The research is performed within the Cognitive Robotics Laboratory. He has published more than 350 contributions across AI, robotics and vision. His research has a strong emphasis on ‘real problems with real solutions’. View full profile
#PressforProgress
International Women's Day. Chief Operating Officer Dr Sue Keay was on the QUT Guild International Women's Day Market & Panel. Image credit: QUT Guild
The 2018 International Women's Day theme is Press for Progress. Centre members talk below about how they will #PressforProgress in 2018
Chief Operating Officer, Dr Sue Keay
I will Press for Progress by celebrating women's achievements in technology by bringing the Grace Hopper Celebration to Australia and New Zealand in 2019. I'm working with the AnitaB.org to make this happen on Wednesday 10th July 2019 in Brisbane, with a view to hosting the celebration (called a Hopperx1) in a different capital city in Australia and NZ every year. Grace Hopper Celebrations are the world's largest gathering of women technologists and have traditionally only been held in the US and, more recently, India. This is a great opportunity to showcase the amazing work done by women technologists in our region and to encourage women to contribute. The purpose of the Grace Hopper Celebration is to "envision a future where the people who imagine and build technology mirror the people and societies for which they build it." 
Centre Advisory Board Member, Kylie Ahern
I know that one person can make an extraordinary impact. And not just politicians or leaders of armies ­– the real change-makers are the men and women who decide to get involved. You don’t need power, just the  determination and willingness to make an effort. 
 
I want to transform how we engage the public in science. I also want to help people understand how they can easily contribute towards reducing the impacts of climate change. We do not have to wait for politicians.  I co-founded and launched COSMOS, and I am currently building an off-grid house that I am publishing a book about. I didn’t know I would achieve these things when I was 30. And I just want to do more.
 
As a publisher and CEO, I have been mentoring young people for two decades. I’ve seen many of my young protégées go on to have amazing achievements of their own.
 
In recent years, my personal philosophy and approach to mentoring have changed. I am much more direct in owning MY successes and achievements, and encouraging other women to do the same. My ‘wake-up’ moment was when a client told me a few years back that I was letting my male colleagues take the public credit for the business, and I seemed to be determined to stand in the shadows, despite being the CEO. That conversation changed me.
 
In 2017 and 2018, we’ve  seen important conversations around women and minority groups, and the structures that hold them back. I will continue to mentor the men and women in my life about how we can encourage more women into science leadership positions.
 
As a species, we cannot afford to squander the brain power, contributions, and perspectives of women as we have in the past. We just don’t have time. It’s that simple.
Centre Director, Distinguished Professor Peter Corke
“The number of women in our field is too low.  I see this reflected in our Centre and in the undergraduate classes that I teach.  This is a big problem for many reasons.  Not utilising half the brainpower on the planet is clearly wasteful, and this fast growing field is going to require an ever growing workforce – we need as many people as we can get.  A subtler, but more insidious problem, is that robotic products will reflect the norms and preferences of those who created them.  A major use case for robots in the future will be human interaction in homes, workplaces, hospitals and schools.  These robots will interact with people of diverse social and cultural backgrounds so having diversity in the design teams is critically important.
 
Nationally we need a major effort to understand why so few women enter engineering and computer science.  Programs to encourage female participation in Australia have run for decades but their effectiveness has to be questioned.  This is clearly not an easy problem to address but there are universities in the US, like Harvey Mudd College, that are able to achieve gender parity in computer science. This means that it is possible.

I’m proud that the Centre is supporting Australia’s first AnitaB.org Grace Hopper Celebration in 2019 and the AnitaB.org has developed a diversity program called BRAID, that we encourage engineering and computer science schools to emulate. Within the Centre we are making some progress.  I am delighted to see more women joining as PhD students and our current research fellow recruitment round is also a great opportunity to increase our diversity. I will press for progress by supporting and mentoring these women by providing opportunities, showcasing their achievements, and working to attract even more female researchers to our Centre ”
ROBOTICS ROADMAP FOR AUSTRALIA
We hope to have the roadmap finished by mid-year and are looking for federal government support to launch the roadmap in June.
 
Until the industry is clearly defined we will struggle with the lack of a clear pipeline for many of our graduates into Australian industry. The roadmap will also address society’s ongoing concerns over fear of job losses, the ethics of automation and how legal frameworks, industry regulations and safety standards need to adapt.
 
To grow Australia’s robotics and computer vision industries we need a strong Tech Sector in Australia. We also need to strengthen our investment in start-ups, and to ensure we have a supportive ecosystem to develop new technologies onshore and build export markets.
 
The focus now turns to ensuring the Roadmap showcases the opportunities that Australia is capable of embracing over the next 15 years.  Australia has the opportunity to take a lead in many sectors where we have pressing needs, e.g., Resources, Agriculture.  We also have niche capabilities in mining automation, providing services in remote regions, and potentially providing field test sites for new large-scale robotic technologies.
 
All key sector workshops have been delivered for the Roadmap, with the Defence workshop taking place in January 2018.  An informative update on the DST Group, together with an overview of the recently announced Defence Cooperative Research Centre (due to open in Queensland later this year) was delivered to over 30 local, and interstate, attendees.
 

Chief Operating Officer Dr Sue Keay
Media enquiries should be directed to Sue Keay 
M. 0408 778 667
E.
sue.keay@roboticvision.org

Project Officer Sandy Holmes
E. sandy.holmes@qut.edu.au
SUMMER SCHOOL SUCCESS
Workshop during RVSS 2018

RVSS WRAPS UP FOR ANOTHER YEAR

This year’s Robotic Vision Summer School continued the trend of recent years and establishes the summer school as one of the premier student events internationally. The summer school is making a home of the picturesque Kioloa coastal campus of ANU and this year the earlier time slot allowed attendees to get a feel for the Australian South East coast in its most beautiful season, luckily without any of the really hot days that can occur.  There was no rain during the week and the evenings around the bonfire talking late into the evening are always a highlight of the summer school. 

The summer school features a varied collection of academic activities. This year the deep dive talks (90 minutes taking you from foundations to state-of-the-art down a narrow path) were given by Yarin Gal who talked about Bayesian deep learning, Margarita Chli who gave a fabulous talk that went from foundations of visual SLAM to applications on aerial vehicles, and Vincent Le Petit who talked about 3D Registration using the now ubiquitous deep learning methods. We had technical presentations from Centre Chief Investigators Chunhua Shen, Matt Dunbabin, Nick Barnes and Tom Drummond, and invited technical talks from Elizabeth Croft (now Associate Investigator with the centre) and Andrea Cherubini who is visiting the QUT node. The tutorial sessions were back in Visual Learning, Visual Semantics in SLAM and Vision and Action and my thanks goes to all the Research Fellows and speakers who contributed to the these; Ravi, Chao, Thanh, Yasir, Vincent , Viorela, Trung, Suman, Valeri, Peter and Andrea.  A special thanks goes to Feras and his team Dimity, Robert, Lachlan, Troy, and Steve for the workshop this year. The workshop is a special part of the summer school and may well be the only time during their PhD that some of the student experience that special feeling of trying to get hardware to work.  It is a feeling that you can’t really understand without having lived through it yourself, and I am proud that there is a whole generation of centre computer vision PhD students who don’t miss out on this formative experience.  This was the first year this particular workshop ran and I appreciate the effort that went into the development of the material.  It was also good to see that most teams made progress but only a couple of teams really got to the point of completing the exercise – life shouldn’t be easy after all.

The summer school had 19 centre PhD students and 31 external students and along with the presenters and organisers we filled the full capacity of the campus accommodation. At the moment we are happy with the size and format of the summer school and I have booked the venue for the next 3 years.  So if you want to come back as a speaker, or you know of someone who we really must invite to come and speak then
get in touch. Overall, the school continues to be a high profile and highly successful activity of the centre.

Professor Rob Mahony
Group bushwalk along the beautiful Pretty Beach to Durrus Mountain walking track
ICRA 2018 IN BRISBANE
62 days to go!

There are only 62 days until the International Conference on Robotics and Automation (ICRA) hits Brisbane. Held at the Brisbane Convention & Exhibition Centre from the 21st to 25th May, ICRA is a premier international forum for robotics researchers to present their work. Centre researchers had 9 papers accepted for the conference and are involved as committee members, presenters and volunteers. The Centre is proud to be a Conference Supporter of ICRA 2018.

Centre Director Peter Corke is a Program Chair.

"The ICRA program is massive, and we expect to welcome over 2000 delegates.  The headline acts are our plenary speakers: Professor Rodney Brooks (Rethink Robotics), Dr Raia Hadsell (Google Deepmind) and Professor Mandyam Srinivasan (Queensland Brain Institute).  There will be 4 tutorials, 31 workshops and 1052 technical papers.  In parallel we have 7 specialist forums, 4 competitions and an exhibition with over 40 companies participating."

Key Centre activity is included below. 


Committee Members
Alex Zelinsky, General Chair
Peter Corke, Program Chair

Rob Mahony, Tutorial Chair
Henrik Christensen, Government Forum Chair
Luis Mejias, Publication Chair
Jon Roberts, Competitions Chair
Thierry Peynot, Local Arrangement Chair

 
See the ICRA 2018 website for full conference details
Forum: Social Robotics
 
Workshop:Long-term Autonomy and Deployment of Intelligent Robots in the Real-world 
Workshop: Advances in Robotic Warehouse Automation
Solutions, Lessons Learned and the Future after the Amazon Robotics Challenge
Workshop: Robotic Vision and Action in Agriculture: the future of agri-food systems and its deployment to the real-world
Tutorial: Vision-based Robot Control 
Tidy Up My Room Challenge
ENVIRONMENT - Where We Live
Epic Duck Challenge. Image credit: J.Hodgson
EPIC DUCK CHALLENGE SHOWS DRONES CAN OUTDO PEOPLE AT SURVEYING WILDLIFE

Centre Research Fellow Trung Pham and Deputy Director Ian Reid have collaborated with colleagues at the University of Adelaide, Monash University and the University of Tasmania for the Epic Duck Challenge. The challenge was developed to see if drones can produce data comparable with traditional wildlife tracking methods. The researchers and volunteers deployed thousands of plastic replica ducks on a beach in Port Willunga, South Australia and tested various methods of tallying them up using drone imagery and human counters

They found that the drones generated accurate wildlife population data with a greater degree of accuracy than the traditional methods using human counters. Drone-derived data were between 43% and 96% more accurate than ground counts, attributed to the drone's superior vantage point in comparison to experienced ground counters.

Trung and Ian along with Jarod Hodgson and Lian Pin Koh from the School of Biological Sciences at the University of Adelaide developed an algorithm and trained it to recognise how the birds appeared in the images. They found that using 10% training data was sufficient to produce a colony count that was comparable to that of a human reviewing the entire scene. This computersation provides the opportunity to cut the costs and resources needed to survey wildlife populations, which could be considerable for sites that are hard to access.


Trung was at Port Willunga for the challenge.
 

“My research is to teach computers to see and understand images. In collaboration with ecologists, I was challenged to develop a computer algorithm to detect and count variable numbers of birds from drone imaging. The task turned out to be very challenging as birds might occupy only a hundred of pixels in the images taken at 100 meters height, unlike cars or pedestrians that usually occupy a large portion of  an image. We were also lacking many “ground-truth” bird templates to train the algorithm with.
 
Our proposed solution only requires a user, given an image, to manually annotate a few bird examples. The algorithm then accurately infers the rest of the image, saving a huge amount of time. The Epic Duck Challenge was a great opportunity to apply computer vision technologies to solve practical real-world problems.”


The article "Drones count wildlife more accurately and precisely than humans" was published in Methods in Ecology and Evolution and featured in The Conversation 
HEALTHCARE - How We Live
Pepper on tour introducing CartMan at the Centre's headquarters at QUT. Image credit: Belinda Ward
ON TOUR WITH PEPPER

Our social robot Pepper (from Softbank Robotics) has become a lot more mobile recently, since we have implemented software to enable the robot to employ traditional 2-dimensional navigation techniques which usually require more expensive scanning lasers. Our technique combines information from Pepper’s depth sensor and point-lasers to navigate within a metric map. We generated an accurate map of the floor at QUT from data collected and processed using CSIRO’s handheld mobile mapping technology (which is commercially available through CSIRO joint-venture company, GeoSLAM Ltd.), and are using that to enable Pepper to move autonomously around the laboratories and office space.
 
Using these new navigation capabilities, Pepper has been conducting tours of the robotics laboratories at QUT, explaining and interacting with our other robots like Harvey the automatic capsicum harvester, and Amazon Robotics Challenge winner, Cartman. One of the first Pepper-led tours was to our project sponsor from the Queensland Government, together with representatives from ST Solutions Australia and (via video conference from Japan) Softbank Robotics.
 
We’re also having success getting Pepper to navigate without a metric map, using purely visual features and deep-learning techniques. This ability will enable Pepper to be deployed in new indoor environments where the robot can be led around and then remember how to reach different destinations using visual cues in a similar way to what a human would.
 
QUT will be hosting a visit and public lecture from Rodolphe Gelin, Chief Scientific Officer for Softbank Robotics Europe, in late March.
Meet the Team behind Pepper
Belinda Ward, Project Manager
Belinda joined the Centre in January 2017 to lead the social robotics project using Softbank's Pepper robot platform. The aim of the project is to understand the role of social robots in influencing human behaviour, and how enhanced robotic vision capabilities can help them interact with humans.
Contact Belinda
 
Suman Bista, Research Fellow
Suman is a Research Fellow working on visual navigation and recognition for Pepper. He completed his PhD titled "Indoor Navigation of Mobile Robots based on Visual Memory and Image-Based Visual Servoing" in 2016 at INRIA.
Contact Suman
Gavin Suddrey, Software Engineer
Gavin is working to expand the general capabilities of the Pepper roboti, primarily through the use of vision. He also assists resaerchers and students within the Centre and wider QUT community to utilise Pepper within their own research.
Contact Gavin 
RESOURCES - Means To Live
Watch the video: Researchers track autonomous underground mining vehicles
EYES UNDERGROUND - A VISION FOR MINE SAFETY
 
Centre Chief Investigator Professor Michael Milford is leading a group of researchers at QUT to develop new technology to equip underground mining vehicles to navigate autonomously through dust, camera blur and bad lighting. Using mathematics and biologically-inspired algorithms, the technology uses vehicle-mounted cameras to track the location of the vehicle in underground tunnels to within metres. The research is a collaboration with Caterpillar, Mining3 and the Queesland Government.
 
Professor Milford said autonomous vehicles operating in underground mines relied on expensive sensing or infrastructure modifications.
 
“This is stage one of the project,” he said. “It’s commercially important to be able to track the location of all your mobile assets in an underground mine, especially if you can do it cheaply without needing to install extra infrastructure or use expensive laser sensing.”
 
“We have developed a positioning system that uses cameras rather than lasers, based on more than a decade of research in biologically-inspired navigation technology.”
 
Read the full media release here
DEVELOPING OUR FUTURE LEADERS
INTERNSHIP AT THE UNIVERSITY OF MARYLAND

Fahimeh Rezazadegan joined the Centre as a PhD researcher in 2015. She recently presented her final seminar for her thesis titled “Human Action recognition and prediction for robotics applications”. During 2017, she did a 5 month internship at the University of Maryland Institute for Advanced Computer Studies (UMIACS) hosted by Professor Larry Davis. Fahimeh talks below about the highlights of the internship.

“I had the chance to visit the UMIACS computer vision group and have brilliant discussions with the researchers there. Almost all the students and research fellows work on deep learning and developing new architectures in different areas from object detection, action recognition, face recognition and so on. We shared our ideas and progress in bi-weekly project meetings with a focus on action recognition.”

During her stay, Fahimeh delivered two talks covering her PhD work and an introductory talk about action prediction in their group and attended the
Workshop on Autonomy, Robotics and Cognition. “I met people in Robotics including Giulio Sandini, Henrik Christensen, Yiannis Aloimonos, Brian Scassellatti, Michael Beetz and many others. It was a great venue to talk about my research and give an overview of our Centre’s research in different domains.”

The trip was a great success – it gave Fahimeh the opportunity to work closely with Professor Davis and this collaboration has resulted in the paper
A Real-time Action Prediction Framework by Encoding Temporal Evolutionsubmitted to the IJCAI conference. “I met people in both the computer vision and robotics fields, showcased our centre’s research, learnt so much via valuable talks and events and have great contacts to keep in touch with for future collaborations. I highly recommend this experience to my fellow PhD researchers”
EDUCATION
A CONVERSATION WITH CENTRE DIRECTOR PETER CORKE
A Conversation with Centre Director Peter Corke. In these video interviews Peter talks about his vision for the Centre, the research challenges its facing and the future of robotic vision.
Click here or on the image to watch
TWO TEAMS SHARE FIRST PLACE SPOT

In a first for the QUT-CSIRO UAV Challenge, teams from two Queensland high schools shared first place in the 2017 challenge. The joint winners were the Cheeky Mangoz from Toowoomba Christian College and the Calamvale Predators from Calamvale Community College. They shared the combined first and second prizemoney in the Queensland Government Airborne Delivery Challenge, receiving $3500 each. 

Now in its 11th year, the UAV Challenge gives young aerial roboticists of the future a chance to think creatively, work in teams and develop the skills they will need to forge a career and lead the industry. Held at Calvert west of Ipswich on September 26-27, the challenge saw 10 teams of students from Australia, the US, Malaysia and Poland use their remotely controlled unmanned aircraft to drop a life-saving medical package to ‘Outback Joe’ who was suffering a severe allergic reaction.

The UAV challenge was co-organised by QUT (Australian Centre for Robotic Vision) and CSIRO’s Data61. Registration for interested High-School teams is now open for the 2018 Challenge.
NEWS, EVENTS AND ACHIEVEMENTS
Congratulations to Centre Director Peter Corke who was named 2017 Australian University Teacher of the Year by the Australian Government's Department of Education and Training in December. 

“Professor Corke has made an outstanding contribution to the physical sciences curriculum with his innovative and engaging teaching of robotics, engaging both his peers and students in Australia and overseas with inventive, high quality robotics education curricula, delivered in the classroom and online,” said Federal Minister for Education and Training Simon Birmingham. Read the full media release
here
Copyright © 2018  Australian Centre for Robotic Vision. All rights reserved.
As you are associated with the Australian Centre for Robotic Vision we have added you to our mailing list.

Our mailing address is:

Centre Headquarters
Australian Centre for Robotic Vision
S Block Level 11
QUT Gardens Point Campus
2 George Street
Brisbane QLD 4001
Australia

Add us to your address book

You can update your preferences or unsubscribe from this list