Membres

Requérents | Principal Investigators

Sabine Süsstrunk

Professeur, EPFL, Laboratoire d’images et représentation visuelle
Full Professor, EPFL, Image and Visual Representation Laboratory

FrançaisEnglish
Raphaël Baroni

Professeur associé, UNIL, Faculté des Lettres, EFLE
Associate Professor, UNIL, Faculty of Arts, EFLE

FrançaisEnglish

Raphaël Baroni est professeur associé à la Faculté des lettres de l’Université de Lausanne, rattaché à l’École de français langue étrangère. Il est président de l’Association internationale des chercheurs en littératures populaires et culture médiatique (LPCM) et membre fondateur, à l’UNIL, du Groupe d’étude sur la bande dessinée(GrEBD) et du Pôle de narratologie transmédiale (NaTrans). Il enseigne depuis plusieurs années un cours de Bachelor d’introduction à l’étude de la bande dessinée et divers cours de Master relatifs à ce médium. Il s’intéresse en particulier à la transition numérique de la culture et en particulier à celle de la bande dessinée.

Baroni est co-dépositaire du projet Sinergia, avec Sabine Süsstrunk et Mathieu Salzmann. Il est responsable de la partie du projet ancrée à l’Université de Lausanne, laquelle vise à retracer l’histoire des reconfigurations de la bande dessinée en lien avec des changements de supports analogiques ou numérique, et à mesurer l’impact des technologies numériques sur la production contemporaine et sur la profession d’auteur ou autrice de bandes dessinées. Il a participé, avec Gaëlle Kovaliv et Olivier Stucky, à la rédaction d’un article de synthèse sur les obstacles qui s’opposent à l’émergence de supports de publication numériques dans le contexte de la bande dessinée européenne et francophone. Il a également collaboré avec Joachim Huguonot à un projet visant à exploiter une approche relevant du « distant viewing » pour étudier l’évolution des formats de publication des magazines pour la jeunesse dans les années 1950-1970. Il collabore aussi avec Bahar Aydemir à une étude empirique de la lecture de planches de bandes dessinées humoristiques francophones utilisant une technologie eye-tracking.

Baroni est notamment l’auteur de La tension narrative (Seuil, 2007), L’œuvre du temps (Seuil, 2009) et Les rouages de l’intrigue (Slatkine , 2017). Il est coéditeur de plusieurs ouvrages ou numéros de revues, parmi lesquels Introduction à l’étude des cultures numériques (Armand Colin, 2020), Narrative Sequence in Contemporary Narratology (Ohio State University Press, 2016) et « Les avatars du chapitre en bande dessinée » (Les Cahiers de narratologie, n° 34, 2018).

Raphaël Baroni is an associate professor at the Faculty of Arts of the University of Lausanne, attached to the School of French as a Foreign Language. He is president of the International Association of Researchers in Popular Literature and Media Culture (LPCM) and a founding member, at the UNIL, of the Study Group on Comics (GrEBD) and of the Pôle de narratologie transmédiale (NaTrans). For several years he has been teaching an introductory Bachelor’s course on the study of comics and various Master’s courses on this medium. He is particularly interested in the digital transition of culture and in particular that of comics.

Baroni is one of the principle investigator of the Sinergia project, together with Sabine Süsstrunk and Mathieu Salzmann. He is responsible for the part of the project based at the University of Lausanne, which aims to retrace the history of the reconfigurations of comics in relation to changes in analogue or digital media, and to measure the impact of digital technologies on contemporary production and on the profession of comics author. He participated, with Gaëlle Kovaliv and Olivier Stucky, in the writing of a synthesis article on the obstacles to the emergence of digital publishing media in the context of European and Francophone comics. He also collaborated with Joachim Huguonot on a project aiming at using a « distant viewing » approach to study the evolution of publication formats of children’s magazines in the 1950s-1970s. He is also collaborating with Bahar Aydemir on an empirical study of the reading of French comic strips using eye-tracking technology.

Baroni is the author of La tension narrative (Seuil, 2007), L’œuvre du temps (Seuil, 2009) and Les rouages de l’intrigue (Slatkine, 2017). He is co-editor of several books or journal issues, including Introduction à l’étude des cultures numériques (Armand Colin, 2020), Narrative Sequence in Contemporary Narratology (Ohio State University Press, 2016) and « Les avatars du chapitre en bande dessinée » (Les Cahiers de narratologie, n° 34, 2018).

Mathieu Salzman

Chercheur senior, EPFL, Laboratoire de vision par ordinateur
Scientist, EPFL, Computer Vision Laboratory

FrançaisEnglish

Mathieu Salzmann est chercheur senior au laboratoire de vision par ordinateur de l’EPFL, avec un « Courtesy Appointment » au Collège des Humanités de l’EPFL. Auparavant, il a été chef de recherche au sein du groupe de vision par ordinateur de NICTA à Canberra, professeur adjoint de recherche à TTI-Chicago et postdoc à ICSI et EECS à UC Berkeley. A l’EPFL, il enseigne des cours de Machine Learning au niveau Bachelor et pour le Master en Humanités Numériques. Ses intérêts de recherche portent sur le développement d’algorithmes d’apprentissage automatique pour des tâches de reconnaissance visuelle, telles que la détection d’objets à base d’images, la segmentation sémantique et la reconstruction 3D. Il a publié plus de 100 articles dans des conférences et revues de premier plan en Computer Vision et en Machine Learning, a été Area Chair pour des conférences de premier plan et est éditeur associé pour IEEE Transactions on Pattern Analysis and Machine Intelligence.

Mathieu Salzmann est co-requérent principal de ce projet Sinergia avec Sabine Süsstrunk et Raphaël Baroni. Dans ce cadre, il supervise Deblina Bhattacharjee, travaillant sur les thèmes du transfert de style entre la bande dessinée et les images naturelles et de l’estimation de la profondeur à partir de cases de bandes dessinées. Avec Sabine Süsstrunk, il co-encadre également Baran Ozadin dans ses travaux sur le « domain adaptation » pour la segmentation sémantique de la bande dessinée. De plus, il a collaboré avec Seungryong Kim et Tong Zhang sur la tâche de reconfiguration d’images, visant à faciliter le transfert de cases de bandes dessinées d’un format à un autre. Il est également impliqué dans la collaboration entre Joachim Hugonot et Raphaël Baroni, visant à un « distant viewing » de la collection de bandes dessinées, numérisée dans le cadre de ce projet, de la Collection Ghebali au Centre BD de la ville de Lausanne.

Mathieu Salzmann is a Senior Researcher at EPFL’s Computer Vision lab, with a courtesy appointment at EPFL’s College of Humanities. Prior to this, he has been a Research Leader in NICTA’s Computer Vision group in Canberra, a Research Assistant Professor at TTI-Chicago, and a postdoctoral fellow at ICSI and EECS at UC Berkeley. At EPFL, he teaches Machine Learning classes at the Bachelor level and for the Master in Digital Humanities. His research interests lie in the development of Machine Learning algorithms for visual recognition tasks, such as image-based object detection, semantic segmentation and 3D reconstruction. He has published over 100 articles at top-tier Computer Vision and Machine Learning conferences and journals, has been an Area Chair for top-tier conferences, and is an Associate Editor of the IEEE Transactions on Pattern Analysis and Machine Intelligence.

Mathieu Salzmann is a Principal Investigator (PI) of this Sinergia project, together with Sabine Süsstrunk and Raphaël Baroni. In this context, he supervises Deblina Bhattacharjee, working on the topics of style transfer between comics and natural images and of depth estimation from comics panels. Together with Sabine Süsstrunk, he also co-supervises Baran Ozadin in his work on domain adaptation for semantic segmentation of comics. Furthermore, he has collaborated with Seungryong Kim and Tong Zhang on the task of image re-targeting, aiming to facilitate the transfer of comic panels from one format to another. He is also involved in the collaboration between Joachim Hugonot and Raphaël Baroni, targeting a distant viewing of the comics collection digitized, in the context of this project, from the Ghebali Collection at Lausanne’s Center for Comics.

Collaborateurs | Collaborators

Zhang Tong

Chercheur PostDoc, EPFL, Laboratoire d’images et de représentation visuelle
PostDoc Researcher, EPFL, Image and Visual Representation Laboratory

English

Tong Zhang is a postdoctoral researcher working in machine learning and computer vision at IVRL Lab, EPFL. He obtained his Ph.D. degree from the Australian National University in 2020. Before that, he received a master degree from New York University and a bachelor degree from Beihang University (BUAA). During his Ph.D. study, he received I received Best Student Paper honorable mention at ACCV 2016 and Paper Award Nominee at CVPR 2020. His research interests include deep geometric learning, clustering, object segmentation, and generative models.

He joined the team in October 2020 working with Prof. Sabine Süsstrunk and Dr. Mathieu Salzmann under the project. Currently, he is working with Deblina Bhattacharjee on multitasks learning estimate depth in comics; with Bahar Aydemir they are working towards saliency detection for studying reader’s reading behavior; with Baran Ozaydin, they are targeting on segmenting comics image with the help from domain adaption. At the same time, he is also focusing on image retargeting and editing to help generate comics objects and images in different resolutions and shapes. He has published two papers on related topics in CVPR 2020 and submitted three papers to conferences. 

Bahar Aydemir

Doctorante, EPFL, Laboratoire d’images et de représentation visuelle
PhD Student, EPFL, Image and Visual Representation Laboratory

English

Bahar Aydemir received B.Sc. degree in Computer Engineering from Middle East Technical University (METU), Ankara, Turkey in 2018. She is currently pursuing the Ph.D. degree in computer and communication sciences under the supervision of Prof. Sabine Süsstrunk, in the Image and Visual Representation Lab (IVRL), École Polytechnique Fédérale de Lausanne (EPFL), Switzerland. She is conducting research on saliency estimation on European comics and natural images. Her current research interests include measuring human attention, computer vision and machine learning.

Within the framework of the project, Bahar is conducting eye-tracking experiments to assess which elements can be modified and even removed in the reconfiguration
process. She collects attention data from various sources such as mouse-tracking, eye-tracking and online crowdsourcing experiments.

As an alternative to collecting attention data from readers, she developed a deep-learning
approach to estimate saliency in comics. Due to the lack of annotated data in the comics domain, this method for now relies on natural images. It uses the high-level and low-level contrast between the multiple objects in a scene. This approach contains an object detector that can be adapted into the comics domain. This study was conducted in collaboration with Deblina Bhattacharjee, Seungryong Kim, and Tong Zhang.

She is also collaborating with Raphaël Baroni on an empirical study of the reading of one-page French comic strips using eye-tracking technology. Bahar participated in the 13th ACM Symposium on Eye Tracking Research and Applications (ETRA2021) between May 24-27 2021.

Deblina Bhattacharjee

Doctorante, EPFL, Laboratoire d’images et de représentation visuelle
PhD Student, EPFL, Image and Visual Representation Laboratory

English

Deblina Bhattacharjee holds a Master’s degree in Computer Science with a specialization in Machine Learning and Optimization from South Korea, 2017. Thereafter, she worked as a deep learning research engineer in Samsung until 2019. She joined EPFL in Lausanne, Switzerland for her doctoral thesis in 2019 where she is working on reconfiguring comics to digital platforms by focusing on depth estimation of comics characters. To achieve this goal, she is studying the depth of comics characters and object instances in the comics panels, so that they can be repositioned. However, estimating depth within such panels is an ill-posed problem as they are monocular and lack annotated ground-truth data. To tackle this, Deblina has explored domain adaptation methods to leverage the depth information from real world images onto the comics domain. In particular, Deblina developed an instance level, detection-based unsupervised image-to-image (I2I) translation technique, which focused on translating the characters in a scene from one domain to another, in this case from the comics domain to the real-world domain. The results of this study led to a publication at the IEEE Conference of Computer Vision and Pattern Recognition (CVPR 2020).

Further, Deblina has developed a contextual saliency method developed in collaboration with Bahar Aydemir to guide her depth estimation task. To that end, Deblina is currently exploring the notion of context between multiple comics panels that will use attention to guide the depth estimation task. Finally, to bring all this together, she is developing an end-to-end multi-task learning architecture that will optimize each task as per the user requirements. This multi-task learning framework will rely on her current findings about context and attention between multiple comics panels. The results are projected to be published in the next conference on computer vision.

Deblina is also an active member of the International Machine Learning Society since 2016 and a Google Women Scholar. She actively volunteers as a mentor for the non-profit organization Teach For India, where she teaches computer science fundamentals and mathematics to high school and undergraduate students, primarily focusing on under-represented factions in STEM fields.

Gaëlle Kovaliv

Doctorante, UNIL, Faculté des Lettres
PhD Student, UNIL, Faculty of Arts

FrançaisEnglish

Diplômée en français moderne et en sciences du langage à l’Université de Lausanne, Gaëlle Kovaliv étudie les conditions de production et de réception des œuvres de bande dessinées nativement numériques francophones. Sa thèse, entre sociologie et lettres, se concentre ainsi sur six axes de recherche : l’identité de ce nouveau médium, les conséquences de la disparition du filtre éditorial – notamment dans une perspective de genrée –, la place du lectorat, l’économie et le financement, le rôle des éditeurs papier et une comparaison internationale des formes de création et de diffusion. Cette recherche est principalement basée sur des entretiens menés auprès de créateurs et créatrices et par des témoignages provenant du monde professionnel de l’édition.

Dans le cadre du projet, Gaëlle Kovaliv a déjà co-rédigé plusieurs articles. Un glossaire bilingue des unités significatives de la bande dessinée avec Olivier Stucky, une étude de la forme des tables des matières des bandes dessinées numériques avec Anaïs Goumand et un article sur le genre, les media studies et la bande dessinée avec Léonore Porchet. Elle a àgalement participé, avec Raphaël Baroni et Olivier Stucky, à la rédaction d’un article de synthèse sur les obstacles qui s’opposent à l’émergence de supports de publication numériques dans le contexte de la bande dessinée européenne et francophone. Elle a aussi réalisé plusieurs bases de données visant à recenser les bandes dessinées nativement numériques francophones ainsi que les créateurs et créatrices actifs et actives dans de domaine.

Par ailleurs, Gaëlle Kovaliv est actuellement co-rédactrice en chef de la revue Comicalités, , membre du Réseau des Narratologues Francophones (RéNaF) et du Pôle de Narratologie Transmédiale (NaTrans). Elle est également co-directrice du festival de bandes dessinées BDFIL .

Gaëlle Kovaliv graduated in modern French and language sciences at the University of Lausanne. She studies the conditions of production and reception of French-speaking digital comics. Her thesis, between sociology and literature, focuses on six main lines of research: the identity of this new medium, the consequences of the disappearance of the editorial filter – particularly from a gendered perspective -, the place of the readership, the economy and financing, the role of paper publishers and an international comparison of the forms of creation and distribution. This research is mainly based on interviews with creators and testimonies from the professional publishing world.

Within the framework of the project, Gaëlle Kovaliv has already co-authored several articles. A bilingual lexicon of significant units in comics with Olivier Stucky, a study of the form of tables of contents in digital comics with Anaïs Goumand and an article on gender, media studies and comics with Léonore Porchet. She participated, with Raphaël Baroni and Olivier Stucky, in the writing of a synthesis article on the obstacles to the emergence of digital publishing media in the context of European and Francophone comics. She also created several databases aiming at identifying French-speaking digital comics and creators active in this field.

In addition, Gaëlle Kovaliv is currently co-editor-in-chief of the journal Comicalités, a member of the Réseau des Narratologues Francophones (RéNaF) and of the Pôle de Narratologie Transmédiale (NaTrans). She the co-director of the comic book festival BDFIL.

Baran Ozaydin

Doctorant, EPFL, Laboratoire d’images et de représentation visuelle
PhD Student, EPFL,Image and Visual Representation Laboratory

English

Baran Ozaydin received B.Sc. degree in Electrical Engineering and a minor degree in Music Performance from Bilkent University, Ankara, Turkey in 2019. He, then, joined Doctoral School of Computer and Communication Sciences at EPFL where he is supervised by Prof. Sabine Süsstrunk and Dr. Mathieu Salzmann. His research is on Domain Adaptation and Image Segmentation. He aims to adapt segmentation models that perform well on natural images to European Comics.

On one hand, Baran works on Image-to-Image translation in order to benefit from the segmentation annotations in natural image datasets. His work aims to improve Image-to-Image translation by guiding the translation with segmentation masks. On the other hand, Baran focuses on merging unsupervised segmentation algorithms with deep learning methods to handle lack of annotated data in comics domain.

Olivier Stucky

Doctorant, UNIL, Faculté des Lettres
PhD Student, UNIL, Faculty of Arts

FrançaisEnglish

Olivier Stucky est titulaire d’un Master en Histoire et esthétique du cinéma et Français moderne. Dans le cadre de sa thèse de doctorat, il étudie les phénomènes de reconfigurations narratives en bande dessinée en s’intéressant aux cas des transferts de récits entre deux supports de publication. Entre théorie et histoire de la bande dessinée, ce travail s’articule autour d’approches systématiques axées sur des cas de transfert intégrés à des stratégies éditoriales à différentes échelles : du périodique illustré vers l’album, de l’album au livre de poche, au CD-ROM ou aux plateformes de lecture en ligne. Le corpus d’analyse mobilisé par Olivier Stucky, fondé sur la production franco-wallonne, comprend ainsi, outre les publications sous forme d’albums qui sont bien connues, des objets aux caractéristiques formelles et matérielles variées, à l’égard desquels la recherche n’a encore consacré que peu de travaux d’ordre théoriques et dont l’histoire éditoriale est souvent méconnue.

Dans le cadre du projet, Olivier Stucky a co-rédigé un glossaire bilingue des unités visuelles et fonctionnelles de la bande dessinée avec Gaëlle Kovaliv et a participé, avec Gaëlle Kovaliv et Raphaël Baroni, à la rédaction d’un article de synthèse sur les obstacles qui s’opposent à l’émergence de supports de publication numériques dans le contexte de la bande dessinée européenne et francophone. Il a également participé à plusieurs colloques internationaux et publiés divers articles scientifiques.

En outre, Olivier Stucky est membre du comité directeur de l’association éditrice des volumes transdisciplinaires Arkhaï dont il a co-édité le volume 2021. Il est également membre du Réseau des Narratologues Francophones (RéNaF), du Pôle de Narratologie Transmédiale (NaTrans) et du GrEBD.

Olivier Stucky holds a Master’s degree in History and Aesthetics of Cinema and Modern French. For his doctoral thesis, he studies the phenomena of narrative reconfigurations in comics by focusing on the cases of narrative transfers between two publication media. Between theory and history of comics, this work is articulated around systematic approaches focused on cases of transfer integrated into editorial strategies at different scales: from the illustrated press to the album, from the album to the pocket book, to the CD-ROM or to online reading platforms. The corpus of analysis mobilised by Olivier Stucky, based on Franco-Belgian production, thus includes, in addition to publications in the well-known form of albums, objects with varied formal and material characteristics, to which the field of comic studies has as yet devoted little theoretical work and whose editorial history is often unknown.

As part of the project, Olivier Stucky co-edited a bilingual lexicon of visual and functional units in comics with Gaëlle Kovaliv and participated, with Gaëlle Kovaliv and Raphaël Baroni, in the writing of a synthesis article on the obstacles to the emergence of digital publishing media in the context of European and Francophone comics. He presented his research at the annual conference of the International Society for the Study of Narratives in Pamplona in 2019. Several other publications and conferences related to his thesis topic are currently on his calendar.

In addition, Olivier Stucky is a member of the steering committee of the transdisciplinary journal Arkhaï and a member of the editorial board of its next issue scheduled for autumn 2021. He is also a member of the Réseau des Narratologues Francophones (RéNaF) and the Pôle de Narratologie Transmédiale (NaTrans).

Peter Grönquist

Informaticien, EPFL, Laboratoire d’images et de représentation visuelle
IT engineer, EPFL
, Image and Visual Representation Laboratory

English

Peter Grönquist is a research software engineer in the Image and Visual Representation Laboratory (IVRL) at EPFL. After a short military career, he obtained his MSc & BSc in computer science at ETH Zürich. During this time he performed research at the Scalable Parallel Computing Laboratory (SPCL) in applying machine- and deep learning to weather forecasting, in a collaboration with the European Center for Medium-Range Weather Forecasting (ECMWF), leading to several publications.

After an industry research internship in computer graphics he recently joined the Sinergia team to support and advance the research through computer vision, machine-/deep learning and software engineering.