BMVA Symposium on Reinforcement Learning in Computer Vision

Chairs: Dmitry Kangin & Nicolas Pugeault (Exeter)

Location: British Computer Society, London

Keynote Speakers:

• Dr Subramanian Ramamoorthy, University of Edinburgh;

• Dr Sen Wang, Heriot-Watt University; and

• Prof Shimon Whiteson, University of Oxford.

• Dr Anil Bharath, Imperial College

Reinforcement learning offers the attractive feature of learning directly from a system’s performance at a task. The fast development of deep learning has allowed to apply reinforcement learning to new classes of problems: playing video games, performing object tracking, predicting eye gaze direction, controlling robots and many others. However, many challenges remain. The high dimensional input spaces require millions of data samples to converge; it prohibits their use for many of the real world, in contrast to the simulated, tasks.

This workshop aims at bringing together experts in machine learning and computer vision to discuss the present and future of reinforcement learning in vision problems.

Programme

09.00 – 9.30: Registration

09.30 – 9.40: Welcome and introduction

09.40 – 10.40: Keynote: Dr Sen Wang, Heriot-Watt University: Learning Robot Localisation and Navigation from Vision

10.40 – 11.10: Sinisa Stekovic, TU Graz: Experience Replay For Learning Multiple Environments.

11.10 – 11.30: Coffee Break

11.30 – 12.30: Keynote: Dr Subramanian Ramamoorthy, Programming by Discussion: New Tools for Human-Robot Interaction

12.30 – 13.30: Lunch break

13.30 – 14.30: Keynote: Prof Shimon Whiteson, Uni of Oxford: Cooperative Multi-Agent Reinforcement Learning

14.30 – 15.00: Dr Dmitry Kangin Exeter: Reinforcement learning and Computer Vision

15.00 – 15.30: Coffee break
15.30 – 16.30: Keynote: Dr Anil Bharath, Imperial: Learning to Act on Visual Data: The Role of Simulation Environments
16.30 – 17.00: Discussion and conclusions