Next Article in Journal
Deep Learning and Handcrafted Features for Virus Image Classification
Previous Article in Journal
Attention-Based Fully Gated CNN-BGRU for Russian Handwritten Text
Article

High-Profile VRU Detection on Resource-Constrained Hardware Using YOLOv3/v4 on BDD100K

1
Instituto Tecnológico de Informática (ITI), Universitat Politècnica de València, 46022 Valencia, Spain
2
Departamento de Informática de Sistemas y Computadores (DISCA), Universitat Politècnica de València, 46022 Valencia, Spain
*
Authors to whom correspondence should be addressed.
Received: 17 November 2020 / Revised: 10 December 2020 / Accepted: 14 December 2020 / Published: 19 December 2020
Vulnerable Road User (VRU) detection is a major application of object detection with the aim of helping reduce accidents in advanced driver-assistance systems and enabling the development of autonomous vehicles. Due to intrinsic complexity present in computer vision and to limitations in processing capacity and bandwidth, this task has not been completely solved nowadays. For these reasons, the well established YOLOv3 net and the new YOLOv4 one are assessed by training them on a huge, recent on-road image dataset (BDD100K), both for VRU and full on-road classes, with a great improvement in terms of detection quality when compared to their MS-COCO-trained generic correspondent models from the authors but with negligible costs in forward pass time. Additionally, some models were retrained when replacing the original Leaky ReLU convolutional activation functions from original YOLO implementation with two cutting-edge activation functions: the self-regularized non-monotonic function (MISH) and its self-gated counterpart (SWISH), with significant improvements with respect to the original activation function detection performance. Additionally, some trials were carried out including recent data augmentation techniques (mosaic and cutmix) and some grid size configurations, with cumulative improvements over the previous results, comprising different performance-throughput trade-offs. View Full-Text
Keywords: on-road detection; artificial intelligence; machine learning; convolutional neural networks; resource-constrained hardware; one-stage detectors; advanced driver-assistance systems; vulnerable road users on-road detection; artificial intelligence; machine learning; convolutional neural networks; resource-constrained hardware; one-stage detectors; advanced driver-assistance systems; vulnerable road users
Show Figures

Figure 1

MDPI and ACS Style

Ortiz Castelló, V.; Salvador Igual, I.; del Tejo Catalá, O.; Perez-Cortes, J.-C. High-Profile VRU Detection on Resource-Constrained Hardware Using YOLOv3/v4 on BDD100K. J. Imaging 2020, 6, 142. https://0-doi-org.brum.beds.ac.uk/10.3390/jimaging6120142

AMA Style

Ortiz Castelló V, Salvador Igual I, del Tejo Catalá O, Perez-Cortes J-C. High-Profile VRU Detection on Resource-Constrained Hardware Using YOLOv3/v4 on BDD100K. Journal of Imaging. 2020; 6(12):142. https://0-doi-org.brum.beds.ac.uk/10.3390/jimaging6120142

Chicago/Turabian Style

Ortiz Castelló, Vicent; Salvador Igual, Ismael; del Tejo Catalá, Omar; Perez-Cortes, Juan-Carlos. 2020. "High-Profile VRU Detection on Resource-Constrained Hardware Using YOLOv3/v4 on BDD100K" J. Imaging 6, no. 12: 142. https://0-doi-org.brum.beds.ac.uk/10.3390/jimaging6120142

Find Other Styles
Note that from the first issue of 2016, MDPI journals use article numbers instead of page numbers. See further details here.

Article Access Map by Country/Region

1
Search more from Scilit
 
Search
Back to TopTop