Dataflow management, dynamic load balancing, and concurrent processing for real‐time embedded vision applications using Quasar

Goossens, Bart

doi:10.1002/cta.2494

Cited by 7 publications

(13 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This creates memory overhead which scales linearly with the number of targets and the number of particles. To mitigate this issue and still retain real-time operation capability, we implemented all tracking and likelihood computation code in CUDA using the high level Quasar compiler and programming language [17]. In our real-world urban environment experiments we measured an average runtime of 52.9ms per frame for the tracking algorithm.…”

Section: Discussionmentioning

confidence: 99%

People Tracking by Cooperative Fusion of RADAR and Camera Sensors

Dimitrievski

Jacobs

Veelaert

et al. 2019

2019 IEEE Intelligent Transportation Systems Conference (ITSC)

View full text Add to dashboard Cite

Accurate 3D tracking of objects from monocular camera poses challenges due to the loss of depth during projection. Although ranging by RADAR has proven effective in highway environments, people tracking remains beyond the capability of single sensor systems. In this paper, we propose a cooperative RADAR-camera fusion method for people tracking on the ground plane. Using average person height, joint detection likelihood is calculated by back-projecting detections from the camera onto the RADAR Range-Azimuth data. Peaks in the joint likelihood, representing candidate targets, are fed into a Particle Filter tracker. Depending on the association outcome, particles are updated using the associated detections (Tracking by Detection), or by sampling the raw likelihood itself (Tracking Before Detection). Utilizing the raw likelihood data has the advantage that lost targets are continuously tracked even if the camera or RADAR signal is below the detection threshold. We show that in single target, uncluttered environments, the proposed method entirely outperforms camera-only tracking. Experiments in a real-world urban environment also confirm that the cooperative fusion tracker produces significantly better estimates, even in difficult and ambiguous situations.

show abstract

Section: Discussionmentioning

confidence: 99%

People Tracking by Cooperative Fusion of RADAR and Camera Sensors

Dimitrievski

Jacobs

Veelaert

et al. 2019

2019 IEEE Intelligent Transportation Systems Conference (ITSC)

View full text Add to dashboard Cite

show abstract

“…A reference area of 340x340 pixels and 117 sections was cropped and used in local template matching for registration. Reconstruction was performed similarly as with the murine heart tissue, using threshold values from the interval [32,78]. As a final step a conversion to 8-bit was always done before exporting to PNG, to allow for final visualisation.…”

Section: Image Restoration Settings For the Experimentsmentioning

confidence: 99%

“…Once the optimal parameters for a specific algorithm are found, the complete data set is ready to be processed. The computationally intensive parts of the workflow are GPU accelerated and indicated with the Quasar logo 32 .…”

Section: Introductionmentioning

confidence: 99%

A “Human-in-the-Loop” Approach for Semi-automated Image Restoration in Electron Microscopy

Roels

Vernaillen²,

Kremer

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

The recent advent of 3D in Electron Microscopy (EM) has allowed for detection of detailed sub-cellular nanometer resolution structures. While being a scientific breakthrough, this has also caused an explosion in dataset size, necessitating the development of automated workflows. Automated workflows typically benefit reproducibility and throughput compared to manual analysis. The risk of automation is that it ignores the expertise of the microscopy user that comes with manual analysis. To mitigate this risk, this paper presents a hybrid paradigm. We propose a 'human-in-the-loop' (HITL) model that combines expert microscopy knowledge with the power of large-scale parallel computing to improve EM image quality through advanced image restoration algorithms. An interactive graphical user interface, publicly available as an ImageJ plugin, was developed to allow biologists to use our framework in an intuitive and user-friendly fashion. We show that this plugin improves visualization of EM ultrastructure and subsequent (semi-)automated segmentation and image analysis. segmentation is done manually 12 , which has several issues. Firstly, such a process is very labor-intensive, making it slow and costly. Secondly, reproducibility is an issue since humans tends to be subjective and biased, and different annotators are likely to produce different segmentations. Lastly, segmentation quality depends on high-quality data, prompting the introduction of quality control.Automated analysis is key to solving the bottleneck in the entire 3D EM workflow. Current advances in large-scale computing and computer vision [13][14][15] can significantly speed up the whole automated image analysis process and make it more cost efficient, potentially leading to challenging and successful research projects that were never possible before 16 . Furthermore, automated methods typically have a formal mathematical formulation and thus the potential to be less biased and more reproducible. Recent automated analysis frameworks are all focused on the segmentation process as such, by either facilitating the manual tracing 17,18 or providing algorithms that allow for interactive learning 19 . As data quality control is omitted, these methods often underperform when challenged with image artifacts such as blur and noise. In practice, this is fairly common in volume EM considering the physical and theoretical issues, including imperfect lenses, thermal heating, diffraction and electron counting errors 20 . Moreover, these artifacts tend to be more pronounced for high throughput data acquisition, which is typically desired in challenging projects 16 . Recent advances in image denoising and deconvolution can significantly mitigate these artifacts 21 .The field of image restoration is a popular field in general image processing: state-of-the-art methods are based on multiresolution shrinkage 22, 23 , non-local pixel averaging 23, 24 , Bayesian estimation 25,26 or convolutional neural networks 27 . Even though most of these methods are available to the communit...

show abstract

“…In “Dataflow Management, Dynamic Load Balancing and Concurrent Processing for Real‐Time Embedded Vision Applications Using Quasar,” B. Goossens explores runtime aspects of Quasar, a high‐level programming language and development environment designed to make the most of underlying heterogeneous hardware in a simple way. Specifically, this paper demonstrates that automatic parallelization and implicit concurrency detection can be achieved in multicore/GPU systems from high‐level code.…”

Section: Smart Camera Hardwarementioning

confidence: 99%