deep neural networks are becoming crucial in many cyber-physical systems involving complex perceptual tasks. For those embedded systems requiring real-time interactions with dynamic environments, as autonomous robots ...
详细信息
deep neural networks are becoming crucial in many cyber-physical systems involving complex perceptual tasks. For those embedded systems requiring real-time interactions with dynamic environments, as autonomous robots and drones, it is of paramount importance that such algorithms are efficiently executed onboard on properly designed hardware accelerators to meet the required performance specifications. In particular, some neural network architectures for object detection and tracking, as You Only Look Once (YOLO), include heavy computational stages that need to be executed before and after the model inference. Such stages are typically not incorporated in traditional accelerators and are executed on general-purpose processors, thus introducing a bottleneck in the overall processing pipeline. To overcome such a problem, this paper presents a generalpurpose accelerator on a field-programmable gate array (FPGA) able to run pre-processing and post-processing operations typically required by vision tasks. The proposed solution has been tested in combination with a YOLO object detector accelerated on an Advanced Micro Devices (AMD) Xilinx Kria KR260 board mounting an UltraScale+ multiprocessor system-on-chip, achieving a significant improvement in terms of both timing performance and power consumption, and enabling onboard visual processing into drones. The proposed solution is able to boost the traditional object detection process by a factor of 4.4, allowing the execution of the full processing pipeline at 60 frames per second (fps), versus 13.6 fps reachable without the proposed accelerator. Asa result, this work enables the use of high-speed cameras for developing more reactive systems that can respond to incoming events with lower latency.
Breakwater construction in Indonesia still relies on divers to direct the placement of rock armour units, which is risky and time-constrained. This research aims to replace the diver's task with a deeplearning-ba...
详细信息
ISBN:
(纸本)9798350392005;9798350391992
Breakwater construction in Indonesia still relies on divers to direct the placement of rock armour units, which is risky and time-constrained. This research aims to replace the diver's task with a deeplearning-based vision system using YOLO-based deeplearning models. The system utilizes image pre-processing technology by applying histogram equalization (HE) techniques to improve image quality before the detection process. This research evaluates the performance of the YOLO-based deeplearning models in detecting armour units in real-time with a focus on various environmental conditions, which are clear and murky water. The analysis reveals clear water consistently supports higher average frame rates (FPS) compared to murky water, maintaining efficient frame processing across all models. In murky water, histogram equalization significantly enhances detection accuracy from 60% to 80% for YOLOv4-tiny and YOLOv7-tiny, demonstrating its effectiveness in challenging conditions. Notably, accuracy remains at 100% for all models in clear water, underscoring their robust performance under optimal visibility conditions.
The safety and durability of concrete structure is an important issue in engineering quality management. In this paper, an imageprocessing algorithm based on deeplearning is proposed to realize real-time quality ins...
详细信息
The safety and durability of concrete structure is an important issue in engineering quality management. In this paper, an imageprocessing algorithm based on deeplearning is proposed to realize real-time quality inspection and automatic defect identification of concrete structures. This algorithm uses the Convolutional Neural Network (CNN) to automatically extract the features of concrete surface quality images, and then identify the existence of defects, thus improving the detection efficiency and accuracy. In this paper, for this method, data samples with different specific structures are collected and manually labeled to the data set;then, a multi-layer CNN model with convolution layer, pool layer and full connection layer is designed to train the model, and then image enhancement technology is used to reduce information noise, and data enhancement technology is used to improve the problem-solving ability of the model. In addition, the strategy of Dropout is used to close some nodes to reduce parameters and prevent over-fitting, and the learning rate is adjusted to optimize the classification effect. In addition, this study constructs an all-weather real-time detection framework, including data acquisition, preprocessing, feature extraction, classification and identification and decision-making alarm system, to ensure the rapid positioning of the detection system. To sum up, the results of this study show that the deeplearningimageprocessing algorithm has good contrast performance in the field of real-time quality inspection of concrete structures. CNN model has better performance than GAN (Generative Adversarial Network) and LSTM (Long Short-Term Memory) models in detection time, defect identification resolution and detection accuracy. The maximum detection time is 366ms and the shortest is 213 ms. The successful development of this algorithm provides a new method for automatic detection of concrete structure quality, which has important application value i
Artificial intelligence (AI) is shaping manufacturing to make it smarter, intelligent, and autonomous. Presently, flexible robots have been introduced that collaborate with humans on the shop floor to enhance producti...
详细信息
Artificial intelligence (AI) is shaping manufacturing to make it smarter, intelligent, and autonomous. Presently, flexible robots have been introduced that collaborate with humans on the shop floor to enhance productivity and efficiency. Object classification and pose estimation in an autonomous robotic system are crucial problems for proper grasping. Extensive research is being conducted to achieve low-cost, computationally efficient, and real-time assessments. However, most of the existing approaches are computationally expensive and constrained to previous knowledge of the 3D structure of an object. This article presents an AI-based solution, which generalizes cuboid- and cylindrical-shaped objects' grasping in real-time, irrespective of the dimensions. The AI algorithm has achieved an average precision of 89.44% and 82.43% for cuboid- and cylindrical-shaped objects. It is identified without the knowledge of the objects' 3D model. The pose is estimated in real-time, accurately. The integrated solution has been implemented in a robotic system fitted with two grippers, a conveyor system, and sensors. Results of several experiments have been reported in this article, which validates the solution. The proposed methodology has achieved 100% accuracy during our experiments to grasp objects on the conveyor belt.
Using big marine data to train deeplearning models is not efficient, or sometimes even possible, on local computers. In this paper, we show how distributed learning in the cloud can help more efficiently process big ...
详细信息
Using big marine data to train deeplearning models is not efficient, or sometimes even possible, on local computers. In this paper, we show how distributed learning in the cloud can help more efficiently process big data and train more accurate deeplearning models. In addition, marine big data is usually communicated over wired networks, which if possible to deploy in the first place, are costly to maintain. Therefore, wireless communications dominantly conducted by acoustic waves in underwater sensor networks, may be considered. However, wireless communication is not feasible for big marine data due to the narrow frequency bandwidth of acoustic waves and the ambient noise. To address this problem, we propose an optimized deeplearning design for low-energy and real-timeimageprocessing at the underwater edge. This leads to trading the need to transmit the large image data, for transmitting only the low-volume results that can be sent over wireless sensor networks. To demonstrate the benefits of our approaches in a real-world application, we perform fish segmentation in underwater videos and draw comparisons against conventional techniques. We show that, when underwater captured images are processed at the collection edge, 4 times speedup can be achieved compared to using a landside server. Furthermore, we demonstrate that deploying a compressed DNN at the edge can save 60% of power compared to a full DNN model. These results promise improved applications of affordable deeplearning in underwater exploration, monitoring, navigation, tracking, disaster prevention, and scientific data collection projects.
The proceedings contain 20 papers. The topics discussed include: edge deployed satellite image classification with TinEViT a X-Cube-AI compatible efficient vision transformer;a design space exploration framework for d...
ISBN:
(纸本)9781510673861
The proceedings contain 20 papers. The topics discussed include: edge deployed satellite image classification with TinEViT a X-Cube-AI compatible efficient vision transformer;a design space exploration framework for deployment of resource-constrained deep neural networks;IoT-enabled unmanned traffic management system with dynamic vision-based drone detection for sense and avoid coordination;exploring action recognition in endoscopy video datasets;integrating image-based LLMs on edge-devices for underwater robotics;age-based clustering of seagrass blades using AI models;layered convolutional neural networks for multi-class image classification;eyeball tracking in closed eyes from shadows;CAEN: efficient adversarial robustness with categorized ensemble of networks;improving real-time security screening;and coupling deep and handcrafted features to assess smile genuineness.
Cloud movement impacts the performance of photovoltaic (PV) power plants by causing sudden fluctuations in output power, leading to voltage instability in connected electricity networks. This paper introduces a novel ...
详细信息
The promotion and utilization of modern technology has made intercultural exchanges more frequent and deep, andsecure image encryption and processing are hot issues for multicultural IOT communication platforms in the...
详细信息
In the rapidly evolving sphere of infrastructure management, early detection of road damage stands paramount for ensuring both safety and longevity. This research introduces an innovative technique for real-time road ...
详细信息
The safety and durability of concrete structure is an important issue in engineering quality management. In this paper, an imageprocessing algorithm based on deeplearning is proposed to realize real-time quality ins...
详细信息
The safety and durability of concrete structure is an important issue in engineering quality management. In this paper, an imageprocessing algorithm based on deeplearning is proposed to realize real-time quality inspection and automatic defect identification of concrete structures. This algorithm uses the Convolutional Neural Network (CNN) to automatically extract the features of concrete surface quality images, and then identify the existence of defects, thus improving the detection efficiency and accuracy. In this paper, for this method, data samples with different specific structures are collected and manually labeled to the data set; then, a multi-layer CNN model with convolution layer, pool layer and full connection layer is designed to train the model, and then image enhancement technology is used to reduce information noise, and data enhancement technology is used to improve the problem-solving ability of the model. In addition, the strategy of Dropout is used to close some nodes to reduce parameters and prevent over-fitting, and the learning rate is adjusted to optimize the classification effect. In addition, this study constructs an all-weather real-time detection framework, including data acquisition, preprocessing, feature extraction, classification and identification and decision-making alarm system, to ensure the rapid positioning of the detection system. To sum up, the results of this study show that the deeplearningimageprocessing algorithm has good contrast performance in the field of real-time quality inspection of concrete structures. CNN model has better performance than GAN (Generative Adversarial Network) and LSTM (Long Short-Term Memory) models in detection time, defect identification resolution and detection accuracy. The maximum detection time is 366ms and the shortest is 213 ms. The successful development of this algorithm provides a new method for automatic detection of concrete structure quality, which has important application value
暂无评论