Kangaroo vehicle collision detection using deep semantic segmentation convolutional neural network
Saleh, Khaled, Hossny, Mohammed and Nahavandi, Saeid 2016, Kangaroo vehicle collision detection using deep semantic segmentation convolutional neural network, in DICTA 2016 : Proceedings of the IEEE International Conference on Digital Image Computing: Techniques and Applications, IEEE, Piscataway, N.J., doi: 10.1109/DICTA.2016.7797057.
Attached Files
Name
Description
MIMEType
Size
Downloads
Title
Kangaroo vehicle collision detection using deep semantic segmentation convolutional neural network
Kangaroo vehicle collisions are a serious problem threatening the safety of the drivers on Australian roads. It is estimated, according to a recent report by Australian Associated Motor Insurers, that there are around 20,000 kangaroo vehicle collisions during year 2015 in Australia. As a result, more than AU $75 million in insurance claims, and a number of animal and human severe injuries and fatalities have been reported. Despite how catastrophic these numbers are, yet a little research has been done in order to avoid or minimise the number of kangaroo vehicle collisions. In this work, we are focusing on the problem of recognising and detecting kangaroos in dynamic environments using a deep semantic segmentation convolutional neural network model. Our model is trained on a synthetic labelled depth images obtained using a simulated range sensor. Our approach records average recall value of over 93% in semantically segmenting any number of kangaroos in the generated testing dataset.
Every reasonable effort has been made to ensure that permission has been obtained for items included in DRO. If you believe that your rights have been infringed by this repository, please contact drosupport@deakin.edu.au.