Springe direkt zu Inhalt

Manshan Guo

Manshan_Photo

Neural Dynamics of Visual Cognition Group

PhD Candidate

Address
Habelschwerdter Allee 45
14195 Berlin

I joined the lab in March 2021 as a doctoral student, co-supervised by Prof. Gemma Roig at Goethe University Frankfurt. Before that I received a BA in Optics (School of Physics, Sun-Yat-Sen University, China) and a MSc in Electronic Engineering (Shanghai institute of applied physics, University of Chinese Academy of Sciences, China).

In my spare time, I enjoy music, dancing, movies, playing Yoga and swimming. 

You can find more details in my CV.

General research interests

My research interest lies in the area where visual cognition and computer vision meet. More particularly, I am interested in applying deep neural networks (DNNs) for brain decoding, investigating whether a DNN trained with brain data (EEG/fMRI) can be transferred to do computer vision tasks, reconstructing visual stimuli from brain data (EEG/fMRI), etc.

I am funded by the Chinese Scholarship Council (CSC). The funding started in March 2021, and will last until March 2025. I currently have one ongoing and two planned research projects.

Ongoing research projects

Project 1: I collaborate with Alessandro Gifford and Martina Vilas and we are interested in detecting how neuroscience can help to improve artificial intelligence (AI). In my case, I first trained the Resnet50 from scratch with Ale’s dataset which includes 16,540 images (Things dataset) paired with corresponding EEG signals for EEG prediction. Then I used this model to synthesize EEG signals to Imagenet datasets (the EEG generated with this method have been proved to be highly correlated with those generated with a linear encoding model). I developed a multi-tasks model based on Resnet50 to do EEG prediction and image classification at the same time. Then we are going to determine how the neural data influences the performance of the DNN in image classification and whether the DNN trained with brain data can perform well in other computer vision tasks.

Student supervision & opportunities 

I am currently not looking for intern/thesis students. However, if your research interests are somewhat overlapping with mine, I am glad to talk to you and let’s see whether we can find a project that we both are interested in.