News

The Project ENNOS integrates color and depth cameras with the capabilities of deep neural networks on a compact FPGA-based platform to create a flexible and powerful optical system with a wide range of applications in production contexts. While FPGAs offer the flexibility to adapt the system to different tasks, they also constrain the size and complexity of the neural networks. The challenge is to transform the large and complex structure of modern neural networks into a small and compact FPGA architecture. To showcase the capabilities of the ENNOS concept three scenarios have been selected. The first scenario covers the automatic anonymization of people during remote diagnosis, the second one addresses semantic 3D scene segmentation for robotic applications and the third one features an assistance system for model identification and stocktaking in large facilities.

During the milestone review a prototype of the ENNOS camera could be presented. It integrates color and depth camera as well as an FPGA for the execution of neural networks in the device. Furthermore, solutions for the three scenarios could be demonstrated successfully with one prototype already running entirely on the ENNOS platform. This demonstrates that the project is on track to achieve its goals and validates the fundamental approach and concept of the project.

Project Partners:
Robert Bosch GmbH
Deutsches Forschungszentrum für Künstliche Intelligenz GmbH (DFKI)
KSB SE & Co. KGaA
ioxp GmbH
ifm eletronic GmbH*
PMD Technologies AG*

*Associated Partner

Contact: Stephan Krauß
Click here to visit our project page.

We are happy to announce that our paper “TGA: Two-level Group Attention for Assembly State Detection” has been accepted for publication at the IEEE International Symposium on Mixed and Augmented Reality (ISMAR), which will take place online from November 9th to 13th. The IEEE ISMAR is the leading international academic conference in the fields of Augmented Reality and Mixed Reality. The symposium is organized and supported by the IEEE Computer Society, IEEE VGTC and ACM SIGGRAPH.

Abstract: Assembly state detection, i.e., object state detection, has a critical meaning in computer vision tasks, especially in AR assisted assembly. Unlike other object detection problems, the visual difference between different object states can be subtle. For the better learning of such subtle appearance difference, we proposed a two-level group attention module (TGA), which consists of inter-group attention and intro-group attention. The relationship between feature groups as well as the representation within a feature group is simultaneously enhanced. We embedded the proposed TGA module in a popular object detector and evaluated it on two new datasets related to object state estimation. The result shows that our proposed attention module outperforms the baseline attention module.

Authors: Hangfan Liu, Yongzhi Su, Jason Raphael Rambach, Alain Pagani, Didier Stricker

Please find our paper here.

Please also check out our YouTube Video.

Contact: Yongzhi.Su@dfki.de, Jason.Rambach@dfki.de

PTC has acquired ioxp GmbH, a German industrial start-up for cognitive AR and AI software. ioxp is a spin-off from the Augmented Vision Department of the German Research Center for Artificial Intelligence GmbH (DFKI). For more Information click here or here (both articles in German only).

Congratulations to Dr. Vladislav Golyanik! He received the DAGM MVTec Dissertation Award 2020 for his outstanding dissertation on “Robust Methods for Dense Monocular Non-Rigid 3DReconstruction and Alignment of PointClouds”. For more Information please click here.

We are happy to announce that our paper “Generative View Synthesis: From Single-view Semantics to Novel-view Images” has been accepted for publication at the Thirty-fourth Conference on Neural Information Processing Systems (NeurIPS 2020), which will take place online from December 6^th to 12^th. NeurIPS is the top conference in the field of Machine Learning. Our paper was accepted from 9454 submissions as one of 1900 (acceptance rate: 20.1%).

Abstract: Content creation, central to applications such as virtual reality, can be a tedious and time-consuming. Recent image synthesis methods simplify this task by offering tools to generate new views from as little as a single input image, or by converting a semantic map into a photorealistic image. We propose to push the envelope further, and introduce Generative View Synthesis (GVS), which can synthesize multiple photorealistic views of a scene given a single semantic map. We show that the sequential application of existing techniques, e.g., semantics-to-image translation followed by monocular view synthesis, fail at capturing the scene’s structure. In contrast, we solve the semantics-to-image translation in concert with the estimation of the 3D layout of the scene, thus producing geometrically consistent novel views that preserve semantic structures. We first lift the input 2D semantic map onto a 3D layered representation of the scene in feature space, thereby preserving the semantic labels of 3D geometric structures. We then project the layered features onto the target views to generate the final novel-view images. We verify the strengths of our method and compare it with several advanced baselines on three different datasets. Our approach also allows for style manipulation and image editing operations, such as the addition or removal of objects, with simple manipulations of the input style images and semantic maps respectively.

Authors: Tewodros Amberbir Habtegebrial, Varun Jampani, Orazio Gallo, Didier Stricker

Please find our paper here.

Please also check out our video on YouTube.

Please contact Didier Stricker for more information.

On July 10th, 2020, Mr Jason Rambach successfully defended his PhD thesis entitled “Learning Priors for Augmented Reality Tracking and Scene Understanding” in front of the examination commission consisting of Prof. Dr. Didier Stricker (TU Kaiserslautern and DFKI), Prof. Dr. Guillaume Moreau (Ecole Centrale de Nantes) and Prof. Dr. Christoph Grimm (TU Kaiserslautern).

In his thesis, Jason Rambach addressed the combination of geometry-based computer vision techniques with machine learning in order to advance the state-of-the-art in tracking and mapping systems for Augmented Reality. His scientific contributions, in the fields of model-based object tracking and SLAM were published in high-rank international peer-reviewed conferences and journals such as IEEE ISMAR and MDPI Computers. His “Augmented Things” paper, proposing the concept of IoT objects that can store and share their AR information received the best poster paper award at the ISMAR 2017 conference.

Jason Rambach holds a Diploma in Computer Engineering from the University of Patras, Greece and a M.Sc. in Information and Communication Engineering from the TU Darmstadt, Germany. Since 2015, he has been at the Augmented Vision group of DFKI where he was responsible for the BMBF-funded research projects ProWiLan and BeGreifen and several industry projects with leading Automotive Companies in Germany. Jason Rambach will remain at DFKI AV as a Team Leader for the newly formed team “Spatial Sensing and Machine Perception” focused on depth sensing devices and scene understanding using Machine Learning.

Patientinnen und Patienten erhalten nach Operationen an Blase, Prostata oder Nieren standardmäßig eine kontinuierliche Dauerspülung der Blase, um Komplikationen durch Blutgerinnsel zu vermeiden. Die Spülung sollte ständig überwacht werden, was jedoch im klinischen Alltag nicht zu leisten ist.

Das Ziel von VisIMon ist es, eine bessere Patientenversorgung bei gleichzeitiger Entlastung des Personals durch eine automatisierte Überwachung der Spülung zu erreichen. Im Projekt wird ein kleines, am Körper getragenes Modul entwickelt, welches den Spülvorgang mit unterschiedlichen Sensoren überwacht. Das System soll sich nahtlos in bestehende Abläufe einfügen lassen. Durch den Zusammenschluss interdisziplinärer Partner aus Industrie und Forschung sollen die notwendigen Sensoren und Schnittstellen entwickelt und zu einem effektiven System vereint werden. Dabei soll moderne Kommunikationstechnologie neue Konzepte ermöglichen, bei denen die Komponenten des Systems drahtlos miteinander kommunizieren, über nutzerfreundliche, interaktive Schnittstellen Daten zur Verfügung stellen und sich durch die Nutzer steuern lassen.

Sensoren, Elektronik zur Auswertung sowie die dazugehörige Systemsoftware zur Bestimmung des Hämoglobins sowie zur Messung der Spülgeschwindigkeit und Füllmengenüberwachung wurden nun erfolgreich am DFKI entwickelt und dem Partner DITABIS zur Integration übergeben. Das System verwendet Eingebettete Künstliche Intelligenz bei der Ermittlung der Messwerte und kann so aktiv und robust auf technische Herausforderungen wie Blasenbildung oder mechanische Erschütterungen reagieren.

Kontakt: Dr. Gerd Reis

Our paper with the title “HandVoxNet: Deep Voxel-Based Network for 3D Hand Shape and Pose Estimation from a Single Depth Map” has been accepted for publication at the IEEE/CVF Conference on Computer Vision and Pattern Recognition 2020 (CVPR 2020) which will take place from June 14th to 19th, 2020 in Seattle, Washington, USA. It is the “premier” conference in the field of Computer Vision. Our paper was accepted from 6656 submissions as one of 1470 (acceptance rate: 22 %).

Abstract 
We propose a novel architecture with 3D convolutions for simultaneous 3D hand shape and pose estimation trained in a weakly-supervised manner. The input to our architecture is a 3D voxelized depth map. For shape estimation, our architecture produces two different hand shape representations. The first is the 3D voxelized grid of the shape which is accurate but does not preserve the mesh topology and the number of mesh vertices. The second representation is the 3D hand surface which is less accurate but does not suffer from the limitations of the first representation. To combine the advantages of these two representations, we register the hand surface to the voxelized hand shape. In extensive experiments, the proposed approach improves over the state-of-the-art for hand shape estimation on the SynHand5M dataset by 47.8%. Moreover, our 3D data augmentation on voxelized depth maps allows to further improve the accuracy of 3D hand pose estimation on real datasets. Our method produces visually more reasonable and realistic hand shapes of NYU and BigHand2.2M datasets compared to the existing approaches.

Please find our paper here.

Authors
Muhammad Jameel Nawaz Malik, Ibrahim Abdelaziz, Ahmed Elhayek, Soshi Shimada, Sk Aziz Ali, Vladislav Golyanik, Christian Theobalt, Didier Stricker

Please also check out our video on YouTube.

Please contact Didier Stricker for more information.

We are very happy to announce that three of our PhD students have been able to successfully defend their PhD thesis during 2019!

Mr. Aditya Tewari defended his thesis with the title “Prior-Knowledge Addition to Spatial and Temporal Classification Models with Demonstration on Hand Shape and Gesture Classification” on October 25th in front of the examination commission consisting of Prof. Dr. Didier Stricker (TU Kaiserslautern and DFKI), Prof. Dr. Paul Lukowicz (TU Kaiserslautern and DFKI) and Prof. Dr. Dr. h. c. Dieter Rombach (Fraunhofer IESE, Kaiserslautern).

Mr. Vladislav Golyanik defended his thesis with the title „Robust Methods for Dense Monocular Non-Rigid 3D Reconstruction and Alignment of Point Clouds” on November 20th in front of the examination commission consisting of Prof. Dr. Didier Stricker (TU Kaiserslautern and DFKI), Prof. Dr. Antonio Aguado (Universitat Politècnica de Catalunya, Spain) and Prof. Dr. Reinhard Koch (Christian-Albrechts-Universität zu Kiel).

Mr. Christian Bailer defended his thesis with the title „New Data Based Matching Strategies for Visual Motion Estimation” on November 22nd in front of the examination commission consisting of Prof. Dr. Didier Stricker (TU Kaiserslautern and DFKI), Prof. Dr. Michael Feslberg (Linköpings University, Sweden) and Dr. Margret Keuper (Max-Planck-Institut für Informatik, Saarbrücken).

All three PhDs have left our Augmented Vision Department shortly after their defense to pursue a career outside of DFKI.

Our team is presenting two papers at the VISAPP 2020 (15th International Conference on Computer Vision Theory and Applications) conference that is taking place from February 27th – 29th in Valletta, Malta.

The two papers are:

Iterative Color Equalization for Increased Applicability of Structured Light Reconstruction
Torben Fetzer, Gerd Reis, Didier Stricker

Autopose: Large-Scale Automotive Driver Head Pose And Gaze Dataset With Deep Head Pose Baseline
Mohamed Selim, Ahmet Firintepe, Alain Pagani, Didier Stricker

The AutoPOSE dataset can be downloaded from the website at autopose.dfki.de.