Welcome!

My name is Zheng-wei Wang (王正蔚). I obtained my Ph.D. at Insight Centre for Data Analytics, Dublin City University, where I was advised by Prof. Tomas E. Ward and Dr. Graham Healy. My Ph.D. is carried out in the area of neuroscience and deep learning. We basically build a system called Neuro-AI Interface for interacting deep neural networks with human neural systems. As the figure shown below, we first use Generative Adversarial Networks (GANs) to produce visual stimulus (i.e., images with different visual quality) to participants, then we can get the human's neural response i.e., EEG in this case for each image (downside path). At the same time, generated images are fed as inputs to a pre-trained deep neural network on the upside path, we also transfer the neural response to the deep neural network. Thus the deep neural network has ability for mapping from images to neural responses like a human brain. We use this framework succesfully to evaluate the performance for different GANs and we name this perceptual score as Neuroscore. More details related to this research can refer to our two papers: Use of neural signals to evaluate the quality of generative adversarial network performance in facial image generation and Synthetic-Neuroscore: Using A Neuro-AI Interface for Evaluating Generative Adversarial Networks.

I am currently a Research Fellow working with Prof. Aljosa Smolic at V-SENSE, Trinity College Dublin. My current research interest focuses on video understanding especially for video action recognition.