Yukiyasu Kamitani (Kyoto University) - Deep Image Reconstruction from the Human Brain

okay did you enjoy the discussion and cake and coffee already okay let's start the final session the next two talks it's related to somehow combination of AI and your science I think and the next speaker is Professor yukia's committin II coming from Tokyo University and he's going to talk about deep image deconstruction from the human

brain completely thank you very much good afternoon I am you coming Danny and I'm here Kota University okay and I joined this she's gone project since the start of the second term 2015 and this term will end soon and I really enjoyed interacting with a group of people and yes I I try to appreciate the mr. son's

initiative and the leadership and partly thanks to this project and activity in this project we were able to have had this distinguished she's come researchers in my department hard coding rusty I say that that's one of the biggest achievement for me okay so my proposal in this project was to build a data-driven approach to study neural

system representation okay so she scan has been studied using very controlled visual stimulus changing parameters or using more diverse expressions but I thought more mid-level representations between fiscal light attributes and more semantic ones very critical components of Cisco by the way which is conquer me many things but in my study I use a siscon to refer to

something that makes an image or is more realistic and to study this mid-level features we used had deep neural networks as the driven feature generators and we compared deep neural network and human brain tivity measured by fMRI and that these are connected by brain decoding so what would render golly this is basically to translate