Non-destructive ultrasonic testing (UT) of materials is used for monitoring critical parts in power plants, aeronautics, oil and gas industry, and space industry. Due to a vast amount of time needed for a human expert...
详细信息
ISBN:
(纸本)9781728131405
Non-destructive ultrasonic testing (UT) of materials is used for monitoring critical parts in power plants, aeronautics, oil and gas industry, and space industry. Due to a vast amount of time needed for a human expert to perform inspection it is practical for a computer to take over that task. Some attempts have been made to produce algorithms for automatic UT scan inspection mainly using older, non-flexible analysis methods. In this paper, two deep learning based methods for flaw detection are presented, YOLO and SSD convolutional neural networks. The methods' performance was tested on a dataset that was acquired by scanning metal blocks containing different types of defects. YOLO achieved average precision (AP) of 89.7% while SSD achieved AP of 84.5%.
It is well known that the sense of smell has significant impacts on human moods, therefore olfactory effects have been widely applied to psychological adjustment as well as clinical treatment. Unlike other senses, sme...
详细信息
ISBN:
(纸本)9781665426565
It is well known that the sense of smell has significant impacts on human moods, therefore olfactory effects have been widely applied to psychological adjustment as well as clinical treatment. Unlike other senses, smell works through the molecules of olfactory stimuli acting on the human nervous system to elicit psychological effects, which is difficult to be accurately described and measured. This makes the commonly used methods hardly applicable to olfactory affective computation. Through analysis of the neural mechanism of human emotions evoked by olfactory sense, this paper specifically designed an EEG experiment to obtain the neural activity data of olfactory stimuli, and compares the clustering characteristics of neural feature data with self-reported scores in PAD emotional space. Thereout, the LS-SVR estimator based on the feature parameters extracted from EEG signal data is proposed for olfactory affective computation. It shows better distinguishing performance and potential reliability than self-reported data, and thus provides an enlightening exploration of this issue.
In recent years, image semantic segmentation based on a convolutional neural network has achieved many advances. However, the development of video semantic segmentation is relatively slow. Directly applying the image ...
详细信息
In recent years, image semantic segmentation based on a convolutional neural network has achieved many advances. However, the development of video semantic segmentation is relatively slow. Directly applying the image segmentation algorithms to each video frame separately may ignore the temporal region continuity inherent in videos. In this study, the authors propose a novel deep neural network architecture with a newly devised spatio-temporal continuity (STC) module for video semantic segmentation. Particularly, the architecture includes an encoding network, an STC module, and a decoding network. The encoding network is used to extract a high-level feature map. The STC module then uses the high-level feature map as input to extract the STC feature map. For decoding, they use four dilated convolutional layers to obtain more abstract representation and a deconvolutional layer to increase the size of the representation. Finally, they fuse the current feature representation and the previous feature representation and get the class probabilities. Thus, this architecture receives a sequence of consecutive video frames and outputs the segmentation result of the current frame. They extensively evaluate the proposed approach on the CamVid and KITTI datasets. Compared with other methods, the authors' approach not only achieves competitive performance but also has lower complexity.
Accelerated stochastic gradient descent (ASGD) methods, which incorporate accelerated proximal gradient (APG) and stochastic gradient (SG), have received considerable attention recently for solving regularized risk mi...
详细信息
Accelerated stochastic gradient descent (ASGD) methods, which incorporate accelerated proximal gradient (APG) and stochastic gradient (SG), have received considerable attention recently for solving regularized risk minimization problems in signal/imageprocessing, statistics and machine learning. However, there has been a paucity of practical guidance proposed for resolving one of the major issues in ASGD: how to choose an appropriate step size. To solve this problem, we propose to use the Barzilai-Borwein (BB) method to automatically compute step size for the accelerated mini-batch Prox-SVRG (Acc-Prox-SVRG) method (the state of the art ASGD method), thereby obtaining a new accelerated method: Acc-Prox-SVRGBB. We prove the convergence of Acc-Prox-SVRG-BB and show that its complexity is comparable with the best known stochastic gradient methods. In addition, we incorporate Beck and Teboulle's APG (FISTA) and Prox-SVRG in a mini-batch setting and obtain another new accelerated gradient descent method, FISTA-Prox-SVRG, which requires the selection of fewer unknown parameters than those required in Acc-Prox-SVRG. Finally, we introduce the BB method into FISTA-Prox-SVRG to further show the efficacy of the BB method. Numerical results demonstrate the advantage of our algorithms. (C) 2019 Elsevier B.V. All rights reserved.
Acoustic scene classification (ASC) and sound event detection (SED) are principal tasks in environmental sound analysis. On the basis of the idea that acoustic scenes and sound events are closely relevant to each othe...
详细信息
ISBN:
(纸本)9781665441629
Acoustic scene classification (ASC) and sound event detection (SED) are principal tasks in environmental sound analysis. On the basis of the idea that acoustic scenes and sound events are closely relevant to each other, some groups previously proposed joint analysis of acoustic scenes and sound events utilizing multitask learning (MTL)-based neural network models. The MTL-based model shares information on acoustic scenes and sound events in mutual estimation. However, in the conventional methods, ASC and SED performances depend strongly on the learning weights of each ASC and SED task, and finding the appropriate balance between the learning weights of ASC and SED tasks is difficult. To address this problem, we therefore propose a dynamic weight adaptation method for multi task learning of ASC and SED based on multi-focal loss in this paper. Experimental results obtained using parts of the TUT Acoustic Scenes 2016/2017 and TUT Sound Events 2016/2017 show that the proposed method improves the scene classification and event detection performance by 3.52 and 3.27 percentage points in micro- Fscore compared with the conventional MTL-based method, respectively. Moreover, the experimental results also indicate that adapting the learning weights dynamically in accordance with the progress of model training improves the ASC and SED performances.
Some industrial processes take place in confined settings only observable by sensors, e.g. infrared (IR) cameras. Drying processes take place while a material is transported by means of a conveyor through a "blac...
详细信息
Superpixel generation is to cluster the pixels with similar features and plays an important role for image segmentation. Conventional superpixel generation methods are more meaningful, however, the learning based meth...
详细信息
ISBN:
(纸本)9781728130385
Superpixel generation is to cluster the pixels with similar features and plays an important role for image segmentation. Conventional superpixel generation methods are more meaningful, however, the learning based method can generate the superpixels directly from the segments in the ground truth and achieve even better performance. In this work, an advanced superpixel generation algorithm that combines the advantages of conventional methods and modern neural network techniques is proposed. In addition to colors and locations, we find that the feature generated by neural networks also provide useful information for superpixel assignment. Simulations show that, with the proposed superpixels, a much more precise segmentation result can be achieved.
Background: In this paper, a Convolutional neural Network (CNN) to learn safe driving behavior and smooth steering manoeuvring, is proposed as an empowerment of autonomous driving technologies. The training data is co...
详细信息
暂无评论