Toward real-time object detection on heterogeneous embedded systems

Niazi-Razavi, Milad; Savadi, Abdorreza; Noori, Hamid

doi:10.1109/iccke48569.2019.8964764

Cited by 5 publications

(3 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Most of them conclude that this protection mechanism significantly reduces the SDCs, but increases the Single Event Functional Interrupts (SEFI) of the algorithms, for example when multiple-bit faults arise. Other authors have evaluated different software hardening strategies, such as Algorithm Based Fault Tolerance (ABFT), Duplication With Comparison (DWC) or Triple Modular Redundancy (TMR) [11], [16].…”

Section: Related Workmentioning

confidence: 99%

Comparison of Parallel Implementation Strategies in GPU-Accelerated System-on-Chip Under Proton Irradiation

Badía

León

Belloch

et al. 2022

IEEE Trans. Nucl. Sci.

View full text Add to dashboard Cite

Commercial Off-The-Shelf (COTS) System-on-Chip (SoC) are becoming widespread in embedded systems. Many of them include a multicore CPU and a high-end GPU. They combine high computational performance with low power consumption and flexible multilevel parallelism. This kind of device is also being considered for radiation environments where large amounts of data must be processed or compute intensive applications must be executed. In this paper we compare three different strategies to perform matrix multiplication in the GPU of a Tegra TK1 SoC. Our aim is to analyze how the different use of the resources of the GPU influences, not only the computational performance of the algorithm, but also its radiation sensitivity. Radiation experiments with protons were performed to compare the behaviour of the three strategies. Experimental results show that most of the errors force a reboot of the platform. The number of errors is directly related with how the algorithms use the internal memories of the GPU, and increases with the matrix size. It is also related with the number of transactions with the global memory, which in our experiments is not affected by the radiation. Results show that the smallest cross-section is obtained with the fastest algorithm, even if it uses the cores of the GPU more intensively.

show abstract

Section: Related Workmentioning

confidence: 99%

Comparison of Parallel Implementation Strategies in GPU-Accelerated System-on-Chip Under Proton Irradiation

Badía

León

Belloch

et al. 2022

IEEE Trans. Nucl. Sci.

View full text Add to dashboard Cite

show abstract

“… The user needs permanent and immediate feedback for reaching the goal and this feedback must result from updated representations of the environment, considering that it can change suddenly (real time processing and feedback)  The destination (static goal) must be defined as coordinates in the 2D occupancy grid like in [24]. We propose, as future work, to compute it by averaging the 3D points obtained from depth data inside a bounding box generated by a module of object detection like [10]- [12].…”

Section: Context and Requirements Of Our Applicationmentioning

confidence: 99%

“…i) These applications require to process dense data at real time, so a computer with high computational capabilities is needed; ii) These applications require that the user moves and carries the hardware, so the computer must also be portable; iii) These applications require complex algorithms to localize in indoors without the help of global positioning system (GPS) (like the ones presented in [7]- [9]), detect static and dynamic obstacles, and build incrementally a representation of the environment; iv) Until few years ago, the algorithms for object detection were not efficient. With the advent in the last years of advanced algorithms of artificial intelligence applied to computer vision (like the ones presented in [10]- [12]), efficient modules for object detection than run on graphics processing unit (GPU) have been plausible; v) Most algorithms for path planning were focused on video games and driverless vehicles (like the ones presented in [13]- [16]), especially due to economic interests. These difficulties have been overcome, due to advances in embedded computers with high capabilities, in vision sensors, and in efficient algorithms of computer vision.…”

Section: Introductionmentioning

confidence: 99%

Path planning for assisting blind people in purposeful navigation

Díaz-Toro

Mosquera-Ortega

Herrera-Silva

et al. 2022

IJEECS

View full text Add to dashboard Cite

Blind <span>people present dificulties for reaching objects of interest in the daily life. In this sense, the integration of a path planning module for assisting blind people in purposeful navigation is noteworthy. In this work, we present an algorithm that leverages the high capability of an embedded computer with graphics processing unit (GPU) NVIDIA Jetson TX2 for computing optimal paths to objects of interest. The algorithm computes the optimal path to the objective, considering changes in the environment and changes in the position of the user. The algorithm is efficient for computing new paths when the environment changes by reusing parts of previous computations. In order to compare the performance, the algorithms were implemented and evaluated in MATLAB, C++ and CUDA, for different size of the grid and percentage of unknown obstacles. We found that the implementation on GPU has a speed up of 20 times W.R.T the implementation in C++ and more than 400 times W.R.T the implementation in MATLAB. These results boost us to integrate this module to our main system based on a stereo camera and a haptic belt and so provide to the user assistance in purposeful navigation at real time.</span>

show abstract