Teilnehmer des Kick-Off-Treffens des KIMBA Forschungsvorhabens stehen vor einem mobilen Prallbrecher von Projektpartner KLEEMANN. // Participants of the kick-off meeting of the KIMBA research project standing in front of a mobile impact crusher from project partner KLEEMANN
[Deutsche Version]
Im Rahmen der Digital GreenTech Konferenz 2023 in Karlsruhe wurden kürzlich 14 neue Forschungsprojekte aus den Bereichen Wasserwirtschaft, nachhaltiges Landmanagement, Ressourceneffizienz und Kreislaufwirtschaft vorgestellt, darunter auch Kimba. Hierbei arbeiten wir gemeinsam mit unseren Projektpartnern an einer KI-basierten Prozesssteuerung und automatisiertem Qualitätsmanagement für das Recycling von Bau- und Abbruchabfällen in Echtzeit. Das spart Kosten, Zeit sowie Ressourcen und schont die Umwelt. So unterstützen wir die Baubranche auf ihrem Weg in die Zukunft.
At the Digital GreenTech Conference 2023 in Karlsruhe, 14 new research projects in the fields of water management, sustainable land management, resource efficiency and circular economy were recently presented, including Kimba. Here, we are working with our project partners on AI-based process control and automated quality management for recycling construction and demolition waste in real time. This saves costs, time and resources and protects the environment. This is how we support the construction industry on its way into the future.
Alt-Text: Teilnehmer des Kick-Off-Treffens des ReVise-UP Forschungsvorhabens stehen vor dem Bergbaugebäude der RWTH Aachen University. // Participants of the kick-off meeting of the ReVise-UP research project stand in front of the mining building of RWTH Aachen University.
Deutsche Version
Forschungsvorhaben „ReVise-UP“ zur Verbesserung der Prozesseffizienz des werkstofflichen Kunststoffrecyclings mittels Sensortechnik gestartet
Im September 2023 startete das vom BMBF geförderte Forschungsvorhaben ReVise-UP („Verbesserung der Prozesseffizienz des werkstofflichen Recyclings von Post-Consumer Kunststoff-Verpackungsabfällen durch intelligentes Stoffstrommanagement – Umsetzungsphase“). In der vierjährigen Umsetzungsphase soll die Transparenz und Effizienz des werkstofflichen Kunststoffrecyclings durch Entwicklung und Demonstration sensorbasierter Stoffstromcharakterisierungsmethoden im großtechnischen Maßstab gesteigert werden.
Auf Basis der durch Sensordaten erzeugten Datentransparenz soll das bisherige Kunststoffrecycling durch drei Effekte verbessert werden: Erstens sollen durch die Datentransparenz positive Anreize für verbesserte Sammel- und Produktqualitäten und damit gesteigerte Rezyklatmengen und -qualitäten geschaffen werden. Zweitens sollen sensorbasiert erfasste Stoffstromcharakteristika dazu genutzt werden, Sortier-, Aufbereitungs- und Kunststoffverarbeitungsprozesse auf schwankende Stoffstromeigenschaften adaptieren zu können. Drittens soll die verbesserte Datenlage eine ganzheitliche ökologische und ökonomische Bewertung der Wertschöpfungskette ermöglichen.
Research project “ReVise-UP” started to improve the process efficiency of mechanical plastics recycling using sensor technology
In September 2023, the BMBF-funded research project ReVise-UP (“Improving the process efficiency of mechanical recycling of post-consumer plastic packaging waste through intelligent material flow management – implementation phase”) started. In the four-year implementation phase, the transparency and efficiency of mechanical plastics recycling is to be increased by developing and demonstrating sensor-based material flow characterization methods on an industrial scale.
Based on the data transparency generated by sensor data, the current plastics recycling shall be improved by three effects: First, data transparency is intended to create positive incentives for improved collection and product qualities and thus increased recyclate quantities and qualities. Second, sensor-based material flow characteristics are to be used to adapt sorting, treatment and plastics processing processes to fluctuating material flow properties. Third, the improved data situation should enable a holistic ecological and economic evaluation of the value chain.
DFKI Augmented Vision researchers Praveen Nathan, Sandeep Inuganti, Yongzhi Su and Jason Rambach received their 1st place award in the prestigious BOP Object Pose Estimation Challenge 2023 in the categories Overall Best RGB Method,Overall Best Segmentation Method and The Best BlenderProc-Trained Segmentation Method.
The BOP benchmark and challenge addresses the problem of 6-degree-of-freedom object pose estimation, which is of great importance for many applications such as robot grasping or augmented reality. This year, the BOP challenge was held within the “8th International Workshop on Recovering 6D Object Pose (R6D)” http://cmp.felk.cvut.cz/sixd/workshop_2023/ at the International Conference on Computer Vision (ICCV) in Paris, France https://iccv2023.thecvf.com/ .
The awards were received by Yongzhi Su and Dr. Jason Rambach on behalf of the DFKI Team and a short presentation of the method followed. The winning method was based on the CVPR 2022 paper “ZebraPose”
The winning approach was developed by a team led by DFKI AV, with contributing researchers from Zhejiang University.
List of contributing researchers:
DFKI Augmented Vision: Praveen Nathan, Sandeep Inuganti, Yongzhi Su, Didier Stricker, Jason Rambach
DFKI Augmented Vision is collaborating with Stellantis on the topic of Radar-Camera Fusion for Automotive Object Detection using Deep Learning. Recently, two new publications were accepted to the GCPR 2023 and EUSIPCO 2023 conferences.
The 2 new publications are:
1. Cross-Dataset Experimental Study of Radar-Camera Fusion in Bird’s-Eye View, Proceedings of the 31st. European Signal Processing Conference (EUSIPCO-2023), September 4-8, Helsinki, Finland, IEEE, 2023.
This paper investigates the influence of the training dataset and transfer learning on camera-radar fusion approaches, showing that while the camera branch needs large and diverse training data, the radar branch benefits more from a high-performance radar.
2. RC-BEVFusion: A Plug-In Module for Radar-Camera Bird’s Eye View Feature Fusion, Proceedings of. Annual Symposium of the German Association for Pattern Recognition (DAGM-2023), September 19-22, Heidelberg, BW, Germany, DAGM, 9/2023.
This paper introduces a new Bird’s Eye view fusion network architecture for camera-radar fusion for 3D object detection that performs favorably on the NuScenes dataset benchmark.
We are happy to announce that the Augmented Vision group will present 4 papers in the upcoming ICCV 2023 Conference, 2-6 October, Paris, France. The IEEE/CVF International Conference in Computer Vision (ICCV) is the premier international computer vision event. Homepage: https://iccv2023.thecvf.com/
The 4 accepted papers are:
U-RED: Unsupervised 3D Shape Retrieval and Deformation for Partial Point Clouds Yan Di, Chenyangguang Zhang, Ruida Zhang, Fabian Manhardt, Yongzhi Su, Jason Raphael Rambach, Didier Stricker, Xiangyang Ji, Federico Tombari
Introducing Language Guidance in Prompt-based Continual Learning Muhammad Gulzain Ali Khan, Muhammad Ferjad Naeem; Luc Van Gool; Federico Tombari; Didier Stricker, Muhammad Zeshan Afzal
DELO: Deep Evidential LiDAR Odometry using Partial Optimal Transport Sk Aziz Ali, Djamila Aouada, Gerd Reis, Didier Stricker
On the 18.6, the team presented their solution and results as part of the workshop program. Scan-to-BIM solutions are of great importance for the construction community as they automate the generation of as-built models of buildings from 3D scans, and can be used for quality monitoring, robotic task planning and XR visualization, among other applications.
Dr. Jason Rambach, coordinator of the EU Horizon Project HumanTech co-organized a workshop on “AI and Robotics in Construction” at the European Robotics Forum 2023 in Odense, Denmark (March 14th to 16th, 2023) in cooperation with the construction Robotics projects Beeyonders and RobetArme.
From the project HumanTech, Jason Rambach presented an overview of the project objectives as well as insights into the results achieved by Month 9 of the project. Patrick Roth from the partner Implenia, presented the perspective and challenges of the construction industry on the use of Robotics and AI in construction sites, while the project partners Dr. Bharath Sankaran (Naska.AI) and Dr. Gabor Sziebig (SINTEF) participated in a panel session discussing the future of Robotics in construction.
We are happy to announce that our article “OPA-3D: Occlusion-Aware Pixel-Wise Aggregation for Monocular 3D Object Detection” was published in the prestigious IEEE Robotics and Automation Letters (RA-L) Journal. The work is a collaboration of DFKI with the TU Munich and Google. The article is openly accessible at: https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=10021668
Abstract: Monocular 3D object detection has recently made a significant leap forward thanks to the use of pre-trained depth estimators for pseudo-LiDAR recovery. Yet, such two-stage methods typically suffer from overfitting and are incapable of explicitly encapsulating the geometric relation between depth and object bounding box. To overcome this limitation, we instead propose to jointly estimate dense scene depth with depth-bounding box residuals and object bounding boxes, allowing a two-stream detection of 3D objects that harnesses both geometry and context information. Thereby, the geometry stream combines visible depth and depth-bounding box residuals to recover the object bounding box via explicit occlusion-aware optimization. In addition, a bounding box based geometry projection scheme is employed in an effort to enhance distance perception. The second stream, named as the Context Stream, directly regresses 3D object location and size. This novel two-stream representation enables us to enforce cross-stream consistency terms, which aligns the outputs of both streams, and further improves the overall performance. Extensive experiments on the public benchmark demonstrate that OPA-3D outperforms state-of-the-art methods on the main Car category, whilst keeping a real-time inference speed.
DFKI Augmented Vision recently released the first publicly available UWB Radar Driving Activity Dataset (RaDA), consisting of over 10k data samples from 10 different participants annotated with 6 driving activities. The dataset was recorded in the DFKI driving simulator environment. For more information and to download the dataset please check the project website: https://projects.dfki.uni-kl.de/rada/
The dataset release is accompanied by an article publication at the Sensors journal:
The Augmented Vision department of DFKI participated in the VIZTA project, coordinated by ST Microelectronics, aiming at developing innovative technologies in the field of optical sensors and laser sources for short to long-range 3D-imaging and to demonstrate their value in several key applications including automotive, security, smart buildings, mobile robotics for smart cities, and industry4.0.
The final project review was successfully completed in Grenoble, France on November 17th-18th, 2022. The schedule included presentations on the achievements of all partners as well as live demonstrators of the developed technologies. DFKI presented their smart building person detection demonstrator based on a top-down view from a Time-of-flight (ToF) camera, developed in cooperation with the project partner IEE. A second demonstrator, showing an in-cabin monitoring system based on a wide-field-of-view, which is installed in DFKIs lab has been presented in a video.
During VIZTA, several key results were obtained at DFKI on the topics of in-car and smart building monitoring including:
7 peer reviewed publications in conferences and journals
DFKI Augmented Vision researchers Yongzhi Su, Praveen Nathan and Jason Rambach received their 1st place award in the prestigious BOP Challenge 2022 in the categories Overall Best Segmentation Method and The Best BlenderProc-Trained Segmentation Method.
The BOP benchmark and challenge addresses the problem of 6-degree-of-freedom object pose estimation, which is of great importance for many applications such as robot grasping or augmented reality. This year, the BOP challenge was held within the “Recovering 6D Object Pose” Workshop at the European Conference on Computer Vision (ECCV) in Tel Aviv, Israel https://eccv2022.ecva.net/ . A total award of $4000 was distributed among the winning teams of the BOP challenge, donated by Meta Reality Labs and Niantic.
The awards were received by Dr. Jason Rambach on behalf of the DFKI Team and a short presentation of the method followed. The winning method was based on the CVPR 2022 paper “ZebraPose”
DFKI Augmented Vision had a strong presence in the recent CVPR 2022 Conference held on June 19th-23rd, 2022, in New Orleans, USA. The IEEE/CVF Computer Vision and Pattern Recognition Conference (CVPR) is the premier annual computer vision event internationally. Homepage: https://cvpr2022.thecvf.com/ .
On June 14th, 2022, Dr. Jason Rambach gave a keynote talk in the Computer Vision session of the Franco-German Research and Innovation Network event held at the Inria headquarters in Versailles, Paris, France. In the talk, an overview of the current activities of the Spatial Sensing and Machine Perception team at DFKI Augmented Vision was presented.
We are happy to announce that the Augmented Vision group will present two papers in the upcoming CVPR 2022 Conference from June 19th-23rd in New Orleans, USA. The IEEE/CVF Computer Vision and Pattern Recognition Conference (CVPR) is the premier annual computer vision event internationally. Homepage: https://cvpr2022.thecvf.com/
Summary: ZebraPose sets a new paradigm on model-based 6DoF object pose estimation by using a binary object surface encoding to train a neural network to predict the locations of model vertices in a coarse to fine manner. ZebraPose shows a major improvement over the state-of-the-art on several datasets of the BOP Object Pose Estimation benchmark.
Summary: We propose a novel Multi-Sphere Image representation called Soft Occlusion MSI (SOMSI) and efficient rendering technique that produces accurate spherical novel-views from a sparse spherical light-field. SOMSI models appearance features in a smaller set (e.g. 3) of occlusion levels instead of larger number (e.g. 64) of MSI spheres. Experiments on both synthetic and real-world spherical light-fields demonstrate that using SOMSI can provide a good balance between accuracy and run-time. SOMSI view synthesis quality is on-par with state-of-the-art models like NeRF, while being 2 orders of magnitude faster.
We are happy to announce that the Augmented Vision group will present 2 papers in the upcoming BMVC 2021 Conference, 22-25 November, 2021:
The British Machine Vision Conference (BMVC) is the British Machine Vision Association (BMVA) annual conference on machine vision, image processing, and pattern recognition. It is one of the major international conferences on computer vision and related areas held in the UK. With increasing popularity and quality, it has established itself as a prestigious event on the vision calendar. Homepage: https://www.bmvc2021.com/
Summary: TICaM is a Time-of-flight In-car Cabin Monitoring dataset for vehicle interior monitoring using a single wide-angle depth camera. The dataset goes beyond currently available in-car cabin datasets in terms of the ambit of labeled classes, recorded scenarios and annotations provided; all at the same time. The dataset is available here: https://vizta-tof.kl.dfki.de/
Summary: Piece-wise 3D planar reconstruction provides holistic scene understanding of man-made environments, especially for indoor scenarios. Different from other existing approaches, we start from enforcing cross-task consistency for our multi-task convolutional neural network, PlaneRecNet, which integrates a single-stage instance segmentation network for piece-wise planar segmentation and a depth decoder to reconstruct the scene from a single RGB image.
DFKI Augmented Vision is working with Stellantis on the topic of Radar-Camera Fusion for Automotive Object Detection using Deep Learning since 2020. The collaboration has already led to two publications, in ICCV 2021 (International Conference on Computer Vision – ERCVAD Workshop on “Embedded and Real-World Computer Vision in Autonomous Driving”) and WACV 2022 (Winter Conference on Applications of Computer Vision).
The 2 publications are:
1. Deployment of Deep Neural Networks for Object Detection on Edge AI Devices with Runtime Optimization, Proceedings of the IEEE International Conference on Computer Vision Workshops – ERCVAD Workshop on Embedded and Real-World Computer Vision in Autonomous Driving
This paper discusses the optimization of neural network based algorithms for object detection based on camera, radar, or lidar data in order to deploy them on an embedded system on a vehicle.
2. Fusion Point Pruning for Optimized 2D Object Detection with Radar-Camera Fusion, Proceedings of the IEEE Winter Conference on Applications of Computer Vision, 2022
This paper introduces fusion point pruning, a new method to optimize the selection of fusion points within the deep learning network architecture for radar-camera fusion.
As part of the research activities of DFKI Augmented Vision in the VIZTA project (https://www.vizta-ecsel.eu/), two publicly available datasets have been released and are available for download. TIMo dataset is a building indoor monitoring dataset for person detection, person counting, and anomaly detection. TICaM dataset is an automotive in-cabin monitoring dataset with a wide field of view for person detection and segmentation and activity recognition. Real and synthetic images are provided allowing for benchmarking of transfer learning algorithms as well. Both datasets are available here https://vizta-tof.kl.dfki.de/. The publication describing the datasets in detail are available as preprints.
