News for March, 2017

Preprint A new paper on Denoising Adversarial Autoencoders is now available on arXiv here.  Code is also available from Toni Creswell on GitHub

Jobs Through collaborations with the DSI, Computing, Aeronautics, the Medical School and Royal College of Music, there are several job opportunities opening up. Please drop us a line at for informal enquiries.


News for February

Rosetrees Trust Award   We have recently received a grant that supports the use of machine learning in cardiovascular risk prediction and treatment planning. This will fund joint research between the ElectroCardioMaths Group, the Aeronautics Department, the Data Sciences Institute and Bioengineering.

DL4I&VU Workshop  We are hosting – together with the Cortexica, the Data Sciences Institute and Neurotechnology CDT – a workshop on Deep Learning for Image and Video Understanding (Feb 24th. 2017).  Future workshops will follow.

News for October

Congratulations to Toni Creswell on her presentation at the VISArt 2016 Conference in Amsterdam. The paper is available on arXiv.

Join us on 3rd November @ Big Data London, 20 Air Street.  AAB will be on the Panel for Deep Learning.

20th October: AAB gave a talk at IDEALondon on Deep Architectures: Past Present and Future.  Slides will soon be available.

News for August

  • Congratulations to Toni Creswell on her paper “Adversarial Training for Sketch Retrieval”, which will appear at an ECCV Workshop; a preprint is available on arXiV.
  • Congratulations to Christoforos Charalambous on his paper “A data augmentation methodology for training machine/deep learning gait recognition algorithms”, at the British Machine Vision Conference.
  • Also this month: “An assistive haptic interface for appearance-based indoor navigation”, is available from here.

Congratulations to Kai & Nat on IJCAI Workshop Paper

Congratulations to Kai Arulkumaran and Nat Dilokthanakul on getting their paper accepted for the 2016 IJCAI Workshop on Deep Reinforcement Learning: Frontiers and Challenges. The paper integrates deep reinforcement learning and hierarchical reinforcement learning by using multiple deep Q-network heads to represent different option policies, with shared convolutional layers to learn common statistical relationships from raw visual inputs. The paper can be viewed on arXiv.

Paper accepted for CVIU

Congratulations to Jose and team for getting to the CVIU Special Issue on computer vision for assistive devices. The paper examines the components of a prototype assistive system for navigation. We look at the accuracy of spatial localization using computer vision, comparing it to the ability of users to sense location via cues provided via a haptic tablet.

Chris Charalambous wins “Best Poster Presentation” at ICDP 2015.

Congratulations to Christoforos Charalambous (BICV Group), who won the prize for “Best Poster Presentation” at the 6th IET Conference on Imaging for Crime Detection and Prevention, held in London, 15-17 July 2015. This study – at the boundary between biomechanics and biometrics – investigates how viewing angle affects the performance of identity recognition using a person’s gait, particularly through the use of joint kinematics. Chris visualised how the accuracy of identity recognition varies with the spatial relationship between a security camera and the trajectory of a person captured using that camera. In addition, he showed how the accuracy of model-based gait recognition depends on the elevation angle of the camera, information which is likely to be useful for companies installing surveillance equipment. Chris’s prize was sponsored by the IEEE.