The Unparalleled Presence of AIML at CVPR 2020

Last week AIML researchers attended CVPR 2020. With Apple, Google and Amazon being among the sponsors, the annual conference is touted as the most revered event for computer vision research. It's said that getting just one paper accepted here is a career highlight, a claim backed up by the submission and acceptance statistics.

AIML had 23 publications accepted in the conference this year, an increase of more than 50% from previous years. These numbers are comparable to those of institutions orders of magnitude larger, and are an incredible result when considering the relative age of AIML.

To have had 23 papers accepted to CVPR is an incredible result, and again puts AIML amongst the world’s best research groups in this very competitive field. CVPR is the highest ranked venue in Computer Vision, and the second highest in all of Engineering and Computer Science.Anton van den Hengel, AIML Director

AIML member Dr Qi Wu, an ARC DECRA fellow at the University of Adelaide School of Computer Science, had six publications accepted this year. REVERIE: Remote Embodied Visual Referring Expression in Real Indoor Environments was one of his team's submissions, and we asked about what this research could lead to.

“You can ask a 10-year-old child to bring you a cushion, and there is a good chance that they will succeed, even in an unfamiliar environment, while the probability that a robot will achieve the same task is significantly lower." Qi says.

“Children have a wealth of knowledge learned from similar environments that they can easily apply, including that cushions may be found on couches, that couches inhabit lounge rooms, and that lounge rooms are often accessed through hallways.”

Qi’s research addresses robots’ lack of these capabilities, and looks to stretch the limits of their domain of application. To be able to ask a multipurpose robot to bring a cushion, or dust a light fitting, or clean up a room may sound like something from science fiction, but thanks to Qi's research, we are now one step closer to making it a reality.

This paper was one of the 335 papers accepted out of the nearly 6,000 submitted in the relevant category — a 5.7% acceptance rate.

These results speak to the talent at the Australian Institute for Machine Learning and we are excited to see our presence increase at CVPR 2021.