ŽáŽèŒ¤‹†ŽÒŠCŠO”hŒ­“ª”]zŠÂƒvƒƒOƒ‰ƒ€E•ñu‰‰
(Reports of Brain Circulation Program)

“úŽžF •½¬29”N2ŒŽ8“ú(…)3ŒÀ (13:30 -- 15:30), 2017/02/8, Wednesday
êŠ(Location)F L1
Ži‰ï(Chair)F ”¨ G–¾(Hideaki HATA)

u‰‰ŽÒ(Presenter)F ’†“‡ —I‘¾(Yuta NAKASHIMA), ‘åã‘åŠwƒf[ƒ^ƒrƒŠƒeƒBƒtƒƒ“ƒeƒBƒA‹@\iInstitute for Datability Science, Osaka Universityj
‘è–Ú(Title)F Report on my visit to Carnegie Mellon University
ŠT—v(Abstract)F Firstly I will briefly introduce CMU and Pittsburgh, where I have been stayed from April 2015 to March 2016. I also talk about what I learned and did there, especially about my research work on visual question answering. Visual question answering is a new task in both computer vision and natural language processing, which tries to answer to a question on an image. We built a new neural net-based model with predicting where to attend in the image to answer a question.

u‰‰ŽÒ(Presenter)F ‘å’J ‚Ü‚ä(Mayu OTANI), Ž‹Šoî•ñƒƒfƒBƒAŒ¤‹†Žº
‘è–Ú(Title)F Activity report at University of Oulu, Finland: Learning joint representations of videos and sentences with web image search
ŠT—v(Abstract)F This report presents research work at Machine Vision Group, University of Oulu, Finland. During my stay, I worked for video retrieval based on natural language queries. Recent work has addressed the problem by embedding visual and textual inputs into a common space where semantic similarities correlate to distances. We propose an embedding approach for sentence and video inputs that combines the advantages of the metadata-based web image search and deep neural network-based representation learning. Overall, we observe an improvement over the state-of-the-art methods in the video and sentence retrieval tasks.

u‰‰ŽÒ(Presenter)F ”–{ Kr(Yukitoshi KASHIMOTO), ƒ†ƒrƒLƒ^ƒXƒRƒ“ƒsƒ…[ƒeƒBƒ“ƒOƒVƒXƒeƒ€Œ¤‹†Žº
‘è–Ú(Title)F Twinkle Megane project and activities in University of Oulu, Finland
ŠT—v(Abstract)F I report my research work and activities in University of Oulu, Finland. In the research project, I worked on the development of Twinkle Megane: Glasses attached with Near-eye LED indicators to support the senior citizen's life.
In addition, I report other life activities in University of Oulu