Alex is a Ph.D. student in the Department of Engineering at the University of Cambridge and a graduate member of Trinity College. Prior to this, Alex studied mechatronics engineering at the University of Auckland where he graduated top of his class. He was awarded a Woolf-Fisher Scholarship to pursue a doctorate at Cambridge. His broad research interests are in artificial intelligence, robotics, control, mechatronics, computer vision and related technologies.
Alex is a member of the Machine Intelligence Lab at Cambridge and is supervised by Prof. Roberto Cipolla. He is interested in deep learning using convolutional neural networks to advance the control and perception of self-driving cars, flying drones and augmented reality.
Alex Kendall, Vijay Badrinarayanan and Roberto Cipolla "Bayesian SegNet: Model Uncertainty in Deep Convolutional Encoder-Decoder Architectures for Scene Understanding." arXiv preprint arXiv:1511.02680, 2015. ( .pdf ) ( bibtex ) ( Webpage )
Vijay Badrinarayanan, Alex Kendall and Roberto Cipolla "SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation." arXiv preprint arXiv:1511.00561, 2015. ( .pdf ) ( poster ) ( bibtex ) ( Webpage )
Alex Kendall and Roberto Cipolla "Modelling Uncertainty in Deep Learning for Camera Relocalization." Proceedings of the International Conference on Robotics and Automation (ICRA), 2016. ( .pdf ) ( bibtex ) ( Webpage )
Alex Kendall, Matthew Grimes and Roberto Cipolla "PoseNet: A Convolutional Network for Real-Time 6-DOF Camera Relocalization." Proceedings of the International Conference on Computer Vision (ICCV), 2015. ( .pdf ) ( poster ) ( bibtex ) ( Webpage )
Kendall, Alex G., Nishaad N. Salvapantula, and Karl A. Stol. "On-board object tracking control of a quadcopter with monocular vision." Unmanned Aircraft Systems (ICUAS), 2014 International Conference on. IEEE, 2014. ( .pdf ) ( bibtex )
SegNet is a deep learning architecture capable of real time semantic segmentation of pixels in an image. The system can segment road, road markings, obstacles, pedestrians and traffic among other important objects to consider while driving. You can view an online demo here.
PoseNet shows we can use convolutional neural networks for robust pose localisation. It uses a deep learning to regress six degree of freedom camera pose relative to a landmark. PoseNet can estimate the camera's location and orientation over large outdoor urban environments or inside buildings. It takes only 5ms to do this from a single colour image. You can view an online demo here.
As a summer research assistant at Auckland University Alex developed a state space model of the Martin Jetpack using systematic frequency analysis techniques.
The Martin Jetpack is a twin co-rotating ducted fan personal aircraft (Martin Jetpack Website). The final model was an 11th order state space representation which was validated with flight test data and used in aircraft simulation and controller design.
For Alex's final year undergraduate research project at Auckland University he developed an autonomous object following control system on a quadcopter using on-board, monocular vision. The quadcopter was able to track trained objects using computer vision processed on-board the quadcopter. The object's x, y coordinate and scale from the on-board video is used to regulate the yaw, height and range respectively. He published this work at the IEEE conference on Unmanned Aircraft Systems 2014. You can view the paper here.
Alex created custom Augmented Reality software to convey design information to clients while working as an engineering consultant at Beca. It was a standalone hardware system using Microsoft Kinect 2.0 and the Unity framework which was able to be deployed for a range of applications.
This was a final year design project in the University of Auckland Mechatronics course. Alex's team's solution was a robot which could autonomously path plan and localise flames (candles), extinguishing them with a CO2 system. The robot was designed and constructed with 3D printing technology. More information is available on the home page.