Intelligence techniques based on convolution neural networks (CNNs) are now dominant in the field of object detection and classification. The deployment of CNNs on embedded edge devices targeting real-time inference sets a challenge due to the limited computing resources and power budgets. Several optimization techniques such as pruning, quantization and use of light neural networks enable the realtime inference but at the cost of precision degradation. However, using efficient approaches to apply the optimization techniques at training and inference stages enable high inference speed with limited degradation of detection performance. In this paper, we revisit the problem of detecting and classifying maritime objects. We investigate different versions of the You Only Look Once (YOLO), a state-of-the-art deep neural network, for real-time object detection and compare their performance for the specific application of detecting maritime objects. The trained YOLO networks are efficiently optimized targeting three recent edge devices: Nvidia Jetson Xavier AGX, AMD-Xilinx Kria KV260 Vision AI Kit, and Movidius Myriad X VPU. The proposed deployments demonstrate promising results with an inference speed of 90 FPS and a limited degradation of 2.4% in mean average precision.
Artificial intelligence (AI) detection techniques based on convolution neural networks (CNNs) require high computations and memory. Their deployment on embedded edge devices, with reduced resources and power budget, is highly hindered especially for applications that requires real-time inference. Several optimization methods such as pruning, quantization and using shallow networks, are mainly utilized to overcome this limitation but at the cost of degradation in detection performance. However, efficient approaches for training and inference have been recently introduced to lower such degradation. This work investigates the use of these approaches to optimize the popular You Only Look Once (YOLO) network targeting various emerging edge devices (Nvidia Jetson Xavier AGX, AMD-Xilinx Kria KV260 Vision AI Kit, and Movidius Myriad X VPU) in order to enhance the detection of humans in maritime environment.
This paper presents new methods for real time estimation of leeway and ocean current, which are based on boat displacements. We propose two solutions that rely on several types of Kalman filters. The first one uses the empirical leeway definition and allows finding the key parameter of this formula. The solution works properly if the error of the formula of leeway remains limited. The second solution takes advantage of an additional sensor and we compare three methods to linearize boat displacements, which are based on a closed-loop model including cascaded filters. These methods are tested on simulation and on real data collected with a maxi multihull. The results first validate the use of a DVL sensor for leeway estimation but also show that it requires the implementation of a complex and specific step of signal processing. Secondly our study demonstrates the relevancy of the closed-loop approach and shows that a solution, based on UKF filters, provides a relevant method to cope with accuracy and stability in case of sensor data outage.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with đź’™ for researchers
Part of the Research Solutions Family.