ANU Computer Science PhD student
Sabre Autonomous Solutions: Engineer
University of NSW
Credit Suisse: Equities Analyst
University of Sydney
NEW! (Sept 2017) We have been selected to receive a Facebook ParlAI research award.
NEW! (26 July 2017) We are 1st in the 2017 Visual Question Answering (VQA) Challenge at CVPR! We are also 1st on the MSCOCO image captioning leaderboard. Details and code are on the project page.
NEW! (July 2017) Our paper on out-of-domain image captioning is accepted to EMNLP 2017.
NEW! (July 2016) We have released a new image caption evaluation metric (SPICE) that improves on CIDEr and METEOR. The paper will appear at ECCV 2016.
Guided Open Vocabulary Image Captioning with Constrained Beam Search
Peter Anderson, Basura Fernando, Mark Johnson and Stephen Gould
In Conference on Empirical Methods for Natural Language Processing (EMNLP), 2017.
An ICP Inspired Inverse Sensor Model with Unknown Data Association
Peter Anderson, Youssef Hunter and Bernhard Hengst
In IEEE International Conference on Robotics and Automation (ICRA), 2013.
Fast Monocular Visual Compass for a Computationally Limited Robot
Peter Anderson and Bernhard Hengst
In Proceedings of the RoboCup International Symposium (RoboCup), 2013.
Robocup Standard Platform League - rUNSWift 2012 Innovations
Sean Harris, Peter Anderson, Belinda Teh, Youssef Hunter, Roger Liu, Bernhard Hengst, Ritwik Roy, Sam Li, Carl Chatfield
In Australasian Conference on Robotics and Automation (ACRA), 2012.
Robot Localisation Using Natural Landmarks
Peter Anderson, Yongki Yusmanthia, Bernhard Hengst and Arcot Sowmya
In Proceedings of the RoboCup International Symposium (RoboCup), 2012.
Three Minute Thesis (3MT) Talk - The Language of Sight. ANU 3MT Finalist, September 2017.
A Practical Introduction to Deep Learning with Caffe. Presented at the Deep Learning Workshop at AI 2015 / ACRA 2015 in December 2015.
FrameFish is an eyewear virtual try-on system for ecommerce websites. I designed it to load and respond more quickly than competing systems, which can take up to 10 seconds to generate a virtual-try-on image (FrameFish takes about 1 second). FrameFish received an Innovate NSW grant and has been featured on Sky Business News and in the UNSW Young Entrepreneurs series.