Modelling Visual Scenes with Psychological Inspiration
Thursday 22 November 2018
Dr Kofi Appiah, Senior Lecturer in the Department of Computing attended the East China Normal University in Shanghai in May 2018 as an invited speaker to give a presentation on his research about how objects in a scene can be modelled with inspiration from human behaviour.
Kofi is experienced in the use of computational intelligence techniques for the analysis and modelling of large video datasets for finding patterns and extracting knowledge. His current research focuses on the interface between embedded computer vision, machine learning and neuroscience; mainly to understand and model the biological vision system on parallel architectures like Field Programmable Gate Arrays (FPGA).
In the presentation, Kofi first briefly introduced the history of artificial intelligence and the rapid development in the past decade, using a simple example of image recognition to give the audience a preliminary understanding of deep neural networks. Then Dr Appiah highlighted the shortcomings of current artificial intelligence in the field of image recognition; that is, artificial intelligence can identify which objects are contained in the image but cannot produce a similar understanding of humans in the scenes presented in the image. Dr Appiah elaborated on the difficulty of the task and the limitations of current artificial intelligence by taking the background and prospects in the separated picture as examples. By introducing experiments on human behaviour in object recognition, Dr Appiah suggested that artificial intelligence should draw inspiration from human behaviour and elaborated on this in conjunction with recent work in his own laboratory.
After the report, the teachers and students who participated in the report had a heated exchange and discussion with Dr Appiah on the comparison between artificial intelligence and human behaviour.