A novel backbone architecture for pedestrian detection based on the human visual system

Saeidi, Mahmoud; Arabsorkhi, Abouzar

doi:10.1007/s00371-021-02280-6

Cited by 14 publications

(7 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several works such as [ 37 , 39 , 45 , 46 ] rely on visual images for detection; [ 37 ] proposed a novel backbone architecture for pedestrian detection based on the human visual system that can be applied to most of the existing architectures. In many cases combining RGB and thermal images has improved the accuracy of object detectors.…”

Section: Related Workmentioning

confidence: 99%

A paced multi-stage block-wise approach for object detection in thermal images

2022

View full text Add to dashboard Cite

The growing advocacy of thermal imagery in applications, such as autonomous vehicles, surveillance, and COVID-19 detection, necessitates accurate object detection frameworks for the thermal domain. Conventional methods could fall short, especially in situations with poor lighting, for instance, detection during night-time. In this paper, we propose a paced multi-stage block-wise framework for effectively detecting objects from thermal images. Our approach utilizes the pre-existing knowledge of deep neural network-based object detectors trained on large-scale natural image data to enhance performance in the thermal domain constructively. The employed, multi-stage approach drives our model to achieve higher accuracies. And the introduction of the pace parameter during domain adaption enables efficient training. Our experimental results demonstrate that the framework outperforms previous benchmarks on the FLIR ADAS dataset on the person, bicycle, and car categories. We have also illustrated further analysis of the framework, such as the effect of its components on accuracy and training efficiency, its generalizability to other thermal datasets, and its superior performance on night-time images in contrast to state-of-the-art RGB object detectors.

show abstract

Section: Related Workmentioning

confidence: 99%

A paced multi-stage block-wise approach for object detection in thermal images

2022

View full text Add to dashboard Cite

show abstract

“…In this case, the database from INRIA (Institut National de Recherche en Informatique et Automatique) [ 45 ] was used, a well known reference for person detection. The results were validated versus the detectors proposed in [ 46 , 47 ]. The NN-specific detection rates are better than those that can be obtained with traditional SVM detectors that extend the HOG model of [ 45 ] with HikSVM (Histogram intersection kernel SVM—trained on the INRIA dataset).…”

Section: Resultsmentioning

confidence: 99%

“…As in [ 46 , 47 ], many references on pedestrian detection focus on 0.1 as the central FPPI for representative MR relevant in comparative performance assessment. As the total number of images in INRIA is 902, FPPI = 0.1 is corresponding, in our case, with FP = 90.2 (as the abscise marked in red on the plot of Figure 10 ), for which our DR = 0.96532, then our representative MR = 3.468%, a good value according to Table 2 .…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Hardware–Software Partitioning for Real-Time Object Detection Using Dynamic Parameter Optimization

Zaharia

Popescu

Sandu

2023

Sensors

View full text Add to dashboard Cite

Computer vision algorithms implementations, especially for real-time applications, are present in a variety of devices that we are currently using (from smartphones or automotive applications to monitoring/security applications) and pose specific challenges, memory bandwidth or energy consumption (e.g., for mobility) being the most notable ones. This paper aims at providing a solution to improve the overall quality of real-time object detection computer vision algorithms using a hybrid hardware–software implementation. To this end, we explore the methods for a proper allocation of algorithm components towards hardware (as IP Cores) and the interfacing between hardware and software. Addressing specific design constraints, the relationship between the above components allows embedded artificial intelligence to select the operating hardware blocks (IP cores)—in the configuration phase—and to dynamically change the parameters of the aggregated hardware resources—in the instantiation phase, similar to the concretization of a class into a software object. The conclusions show the benefits of using hybrid hardware–software implementations, as well as major gains from using IP Cores, managed by artificial intelligence, for an object detection use-case, implemented on a FPGA demonstrator built around a Xilinx Zynq-7000 SoC Mini-ITX sub-system.

show abstract

“…In order to ensure complete coverage of the zone of operation of the RSU, it is necessary to place them at a distance of approximately 12 meters. The task of RSU enriched with PIR sensors is to provide information about the presence of wildlife [24]. In the RSU, data is collected and, if necessary, local data processing is performed, and then the relevant data is sent via a communication device to the central processing and data storage in the cloud, Point 7 and Point 6 in Fig.…”

Section: Model Of Vehicle-wildlife Collision Avoidance System On Rura...mentioning

confidence: 99%

Modeling a LoRAWAN Network for Vehicle Wildlife Collision Avoidance System on Rural Roads

Jotanovic,

Jausevac,

Perakovic

et al. 2024

Preprint

View full text Add to dashboard Cite

The network of rural roads covers different types of terrain, including forest areas, pastures, arable land and sparsely populated areas. The safety of people and animals is a priority in traffic on these roads. Early detection of pedestrians, animals and other moving objects along the road can significantly reduce the risk of accidents. As part of this research, a sensor system is being developed that can detect characteristics of living things in motion, such as unexpectedly crossing the road without clear signs. Such timely detection of moving objects enables adequate preventive measures and reduces potential traffic accidents. The consequences of traffic accidents of this type can cause serious damage to animals and people property, and road infrastructure. The topicality of this problem at the spatial and seasonal level is emphasized in studies that identify the hotspots of these accidents. Factors such as traffic characteristics and road infrastructure are key to modeling protective systems on rural roads. The presented study investigates the deployment of sensor nodes and LoRAWAN gateways for wildlife detection on rural roads, with the aim of reducing the risk of traffic accidents caused by Wildlife-Vehicle Collisions.

show abstract

A novel backbone architecture for pedestrian detection based on the human visual system

Cited by 14 publications

References 46 publications

A paced multi-stage block-wise approach for object detection in thermal images

A paced multi-stage block-wise approach for object detection in thermal images

Hardware–Software Partitioning for Real-Time Object Detection Using Dynamic Parameter Optimization

Modeling a LoRAWAN Network for Vehicle Wildlife Collision Avoidance System on Rural Roads

Contact Info

Product

Resources

About