Skip to main content

Depth of field for machine vision

For his outstanding doctoral thesis on CAD-based visual recognition, computer scientist Dr. Benjamin Planche has been awarded the Dissertation Award 2022 by the Franco-German University (FGU). He wrote the thesis within the framework of a doctoral track programme offered jointly by the University of Passau and INSA Lyon in France.

The award was presented at the end of January at a formal virtual ceremony attended by the ambassador of the Federal Republic of Germany to France, Dr. Hans-Dieter Lucas, and the president of the Franco-German University (FGU), Professor Philippe Gréciano. The FGU initially established the four dissertation awards to recognise outstanding research conducted by young researchers undertaking Franco-German doctoral studies. Dr. Benjamin Planche was presented the FGU dissertation award worth € 4,500, while three others doctoral graduates were given the "Prix européen", the "Prix international" and the "Prix des affaires", each of which comes with prize money of € 2,000.

Schwarz-Weißes synthetisches Tiefenbild, das mit der in Benjamin Planches Diplomarbeit beschriebenen Methode (DepthSynth) erstellt wurde

While the human eye can only detect gray colored areas in this image, the computational eye can extract important information for industrial production from the image: Example of a synthetic depth image of an industrial engine created using the method described in Benjamin Planche's thesis (DepthSynth). Photo: Benjamin Planche.

In his doctoral thesis entitled "Bridging the Realism Gap for CAD‐Based Visual Recognition" – which Dr. Planche submitted in 2020 earning him the highest possible distinction of "summa cum laude" – he developed three innovative techniques to generate synthetic image data, and these have already successfully found their way into industrial use. His first paper introduces a new method of rendering realistic depth images from 3D models which is able to fully simulate the output of modern depth cameras. Where no relevant 3D models for the production of synthetic images are available, Dr. Planche suggests a method that effectively builds a representation of visual scenes from partial observations. His first two research papers focus on bringing synthetic image data closer to the real time domain, whereas the third paper implements the opposite process by mapping unseen target images.

"The results of the research conducted by Benjamin Planche are a significant contribution to model-based quality testing of building components in industrial manufacture and maintenance," says Professor Harald Kosch, who holds the chair for Distributed Information Systems at the University of Passau and served as first referee of the thesis. "I'm particularly pleased that his excellent thesis has been given due recognition in the form of the FGU Dissertation Award. What's more, this distinction is proof of the splendid and successful academic exchange happening between Passau and Lyon and the close partnership between the two universities."

Example of how depth images work: The image on the right shows the color image that conforms to our viewing habits. The image on the left shows the corresponding depth image. These kinds of images are used in industrial production and manufacturing control:

Schwarz-weißes Tiefenbild auf welchem die Umrisse der Gegenstände des Farbfotos erkennbar sind

Photo: Licence Creative Commons Attribution 4.0 International. Authors: Hodaň, P. Haluza, Š. Obdržálek, J. Matas, M. Lourakis, X. Zabulis, T-LESS: An RGB-D Dataset for 6D Pose Estimation of Texture-less Objects, WACV 2017. Available here.

Farbfoto zu welchem das Tiefenbild generiert wurde. Zu sehen sind verschiedene Gegenstände, wie Klebeband und eine Taschenlampe.

Photo: Licence Creative Commons Attribution 4.0 International. Authors: Hodaň, P. Haluza, Š. Obdržálek, J. Matas, M. Lourakis, X. Zabulis, T-LESS: An RGB-D Dataset for 6D Pose Estimation of Texture-less Objects, WACV 2017. Available here.

The doctoral thesis was written in the framework of the Franco-German graduate research training group "Multimedia, Distributed and Pervasive Secure Systems“ (MDPS) between the University of Passau and INSA Lyon. This research training group offers doctoral students of Computer Science at the University of Passau, INSA Lyon and the University of Milan the possibility of completing a doctoral programme under the supervision of two universities (cotutelle de thèse). In addition to cross-country supervision, the MDPS also offers biannual workshops with seminars and guest lectures held by experts from industry and research where doctoral students receive valuable feedback and are given the opportunity to engage in an exchange about their current research progress.

The papers of the doctoral thesis have appeared in several different academic publications and presented on various occasions, including at the renowned Conference on Neural Information Processing Systems. Dr. Planche had previously been awarded the Dissertation Award of Sparda-Bank Ostbayern eG at the Dies Academicus 2021 of the University of Passau.

Personal profile

Dr. Benjamin Planche was born in Moûtiers, France, in 1990. After graduating from Lycée Jeanne d’Arc in Albertville, France, he set out to study computer science at INSA Lyon and was accepted onto the double master's programme offered in cooperation with the University of Passau in 2013. During his studies, he spent one semester at Luleå University of Technology in Sweden. He completed his studies with distinction in 2014, starting out on his doctoral studies in computer science at the University of Passau one year later. At the same time, Siemens AG in Munich awarded him a doctoral scholarship. After finishing his doctoral thesis with "summa cum laude" in 2020, he took a postdoc position at Siemens Technology AG in Princeton, USA, where he did research on robust visual recognition systems. Currently, Dr. Planche is employed as a Senior Research Scientist at UII America, Inc. There, he and his colleagues are working on improving and automating medical imaging workflows.

About the Franco-German University

The Franco-German University (FGU) is an institution founded in 1997 by France and Germany with the task of promoting binational double degree study programmes as well as the binational education of doctoral students and young researcher mobility between the two countries. The FGU network includes over 200 universities, universities of applies sciences and grandes écoles in around 130 German and French cities. Around 6,400 students and 150 doctoral students are presently enrolled in the cooperation programmes sponsored by the FGU.

To learn more about the FGU and its awards for excellence, check out the website.

Professor Harald Kosch

Professor Harald Kosch

conducts research on distributed and multimedia systems

How can intelligent systems source information from across various media?

How can intelligent systems source information from across various media?

Professor Harald Kosch has held the Chair of Distributed Information Systems at the Faculty of Computer Science and Mathematics since 2006. He is the German director of DFH/UFA Doctoral College, which brings together INSA de Lyon and the University of Passau. The main purpose of this exchange is to facilitate cooperation projects in intelligent digital systems.

Playing the video will send your IP address to an external server.

Show video