Jeudi 28 avril à 11h00 salle Aurigny
Première session de "LinkMedia Speaks Science", le rendez-vous scientifique bi-hebdomadaire proposé par l'équipe LinkMedia.
Cette première édition accueillera Gül Varol , chercheuse permanente dans l'équipe IMAGINE de l'École des Ponts ParisTech pour un exposé intitulé "What can we learn from subtitled sign language data? ".
Abstract : Research on sign language technologies has suffered from the lack of data to train machine learning models. This talk will describe our recent efforts on scalable approaches to automatically annotate continuous sign language videos with the goal of building a large-scale dataset. In particular, we leverage weakly-aligned subtitles from sign interpreted broadcast footage. These subtitles provide us with candidate keywords to search and localise individual signs. To this end, we develop three sign spotting techniques (i) using mouthing cues at the lip region, (ii) looking up videos from sign language dictionaries, and (iii) exploring the sign localisation that emerges from the attention mechanism of a sequence prediction model. With these methods, we build the BBC-Oxford British Sign Language Dataset (BOBSL), continuous signing videos of more than a thousand hours, containing millions of sign instance annotations from a large vocabulary. More information about the dataset can be found at https://arxiv.org/abs/2111.03635.
Gül Varol est chercheuse permanente dans l'équipe IMAGINE de l'École des Ponts ParisTech. Auparavant, elle était chercheur postdoctoral à l'Université d'Oxford (VGG), travaillant avec Andrew Zisserman. Elle a obtenu son doctorat dans l'équipe WILLOW de l'Inria Paris et de l'École Normale Supérieure (ENS). Sa thèse, co-encadrée par Ivan Laptev et Cordelia Schmid, a reçu les prix de thèse d'ELLIS et d'AFRIF. Au cours de son doctorat, elle a passé du temps à MPI, Adobe et Google. Avant cela, elle a obtenu un baccalauréat et une maîtrise à l'Université de Boğaziçi. Ses recherches portent sur la vision par ordinateur, en particulier l'apprentissage de la représentation vidéo, l'analyse du mouvement humain et les langues des signes.
Thursday 28 April at 11am in the Aurigny Room
First session of "LinkMedia Speaks Science", the bi-weekly guest speakers proposed by the LinkMedia team
This first edition will welcome Gül Varol , permanent researcher in the IMAGINE team at École des Ponts ParisTech for a talk entitled "What can we learn from subtitled sign language data? ".
Abstract : Research on sign language technologies has suffered from the lack of data to train machine learning models. This talk will describe our recent efforts on scalable approaches to automatically annotate continuous sign language videos with the goal of building a large-scale dataset. In particular, we leverage weakly-aligned subtitles from sign interpreted broadcast footage. These subtitles provide us with candidate keywords to search and localise individual signs. To this end, we develop three sign spotting techniques (i) using mouthing cues at the lip region, (ii) looking up videos from sign language dictionaries, and (iii) exploring the sign localisation that emerges from the attention mechanism of a sequence prediction model. With these methods, we build the BBC-Oxford British Sign Language Dataset (BOBSL), continuous signing videos of more than a thousand hours, containing millions of sign instance annotations from a large vocabulary. More information about the dataset can be found at https://arxiv.org/abs/2111.03635.
Gül Varol is a permanent researcher in the IMAGINE team at École des Ponts ParisTech. Previously, she was a postdoctoral researcher at the University of Oxford (VGG), working with Andrew Zisserman. She obtained her PhD from the WILLOW team of Inria Paris and École Normale Supérieure (ENS). Her thesis, co-advised by Ivan Laptev and Cordelia Schmid, received the PhD awards from ELLIS and AFRIF. During her PhD, she spent time at MPI, Adobe, and Google. Prior to that, she received her BS and MS degrees from Boğaziçi University. Her research is focused on computer vision, specifically video representation learning, human motion analysis, and sign languages.