Search results

The paper aims to build a new objective evaluation method of fabric pilling by combining an integrated image analysis technology with a deep learning algorithm.

Series of image analysis techniques were adopted. First, a Fourier transform transformed images into the frequency domain. The optimal resolution matrix of an exponential high-pass filter was determined by combining the energy algorithm. Second, the multidimensional discrete wavelet transform determined the optimal division level. Third, the iterative threshold method was used to enhance images to obtain a complete and clear pilling ball images. Finally, the deep learning algorithm was adopted to train data from pilling ball images, and the pilling levels were classified according to the learning features.

The paper provides a new insight about how to objectively evaluate fabric pilling grades. Results of the experiment indicate that the proposed objective evaluation method can obtain clear and complete pilling information and the classification accuracy rate of the deep learning algorithm is 94.2%, whose structures are rectified linear unit (ReLU) activation function, four hidden layers, cross-entropy learning rules and the regularization method.

Because the methodology of the paper is based on woven fabric, the research study’s results may lack generalizability. Therefore, researchers are encouraged to test other kinds of fabric further, such as knitted and unwoven fabrics.

Combined with a series of image analysis technology, the integrated method can effectively extract clear and complete pilling information from pilled fabrics. Pilling grades can be classified by the deep learning algorithm with learning pilling information.

Indian food recognition can be considered as a case of fine-grained type visual recognition, where the several photos of same category generally have significant variability. Therefore, effective segmentation and classification technique is required to identify the particular cuisines and fine-grained analysis. The paper aims to discuss this issue.

In this paper, the authors provided an effective segmentation approach through the proposed edge adaptive (EA)-deep convolutional neural networks (DCNNs) model, where each input images are divided into patches in order to provide much efficient and accurate structural description of data.

EA-DCNNs starts with developing a coarse map of feature that obtained through DCNN, afterwards EA model is applied to construct the final segmented image.

The training model of EA-DCNN consists of pooling, rectified linear unit and convolution, which help convolutional network to optimize the performance of segmentation in a significant extent, which is much practical and relevant in the context of food image segmentation.

The purpose of this paper is to improve the precision of gangue detection. In the real production environment, some gangue features are not obvious, and it is difficult to distinguish between coal and gangue. The color of the conveyor belt is similar to the gangue, the background noise also brings challenge to gangue detection. To address the above problems, we propose a feature aggregation method based on optical flow (FAOF).

An FAOF is proposed. First, to enhance the feature representation of the current frame, FAOF applies the timing information of video stream, propagates the feature information of the past few frames to the current frame by optical flow. Second, the coordinate attention (CA) module is adopted to suppress the noise impact brought by the background of convey belt. Third, the Mish activation function is used to replace rectified linear unit to improve the generalization capability of our model.

The experimental results show that the gangue detection model proposed in this paper improve 4.3 average precision compared to baseline. This model can effectively improve the accuracy of gangue detection in real production environment.

The key contributions are as follows: this study proposes an FAOF; this study adds CA module and Mish to reduce noise from the background of the conveyor belt; and this study also constructs a large gangue data set.

The expansion of marketing data is encouraging the growing use of deep learning (DL) in marketing. I summarize the intuition behind deep learning and explain the mechanisms of six popular algorithms: three discriminative (convolutional neural network (CNN), recurrent neural network (RNN), and Transformer), two generative (variational autoencoder (VAE) and generative adversarial networks (GAN)), and one RL (DQN). I discuss what marketing problems DL is useful for and what fueled its growth in recent years. I emphasize the power and flexibility of DL for modeling unstructured data when formal theories and knowledge are absent. I also describe future research directions.

The study proposed a human–robot interaction (HRI) framework to enable operators to communicate remotely with robots in a simple and intuitive way. The study focused on the situation when operators with no programming skills have to accomplish teleoperated tasks dealing with randomly localized different-sized objects in an unstructured environment. The purpose of this study is to reduce stress on operators, increase accuracy and reduce the time of task accomplishment. The special application of the proposed system is in the radioactive isotope production factories. The following approach combined the reactivity of the operator’s direct control with the powerful tools of vision-based object classification and localization.

Perceptive real-time gesture control predicated on a Kinect sensor is formulated by information fusion between human intuitiveness and an augmented reality-based vision algorithm. Objects are localized using a developed feature-based vision algorithm, where the homography is estimated and Perspective-n-Point problem is solved. The 3D object position and orientation are stored in the robot end-effector memory for the last mission adjusting and waiting for a gesture control signal to autonomously pick/place an object. Object classification process is done using a one-shot Siamese neural network (NN) to train a proposed deep NN; other well-known models are also used in a comparison. The system was contextualized in one of the nuclear industry applications: radioactive isotope production and its validation were performed through a user study where 10 participants of different backgrounds are involved.

The system was contextualized in one of the nuclear industry applications: radioactive isotope production and its validation were performed through a user study where 10 participants of different backgrounds are involved. The results revealed the effectiveness of the proposed teleoperation system and demonstrate its potential for use by robotics non-experienced users to effectively accomplish remote robot tasks.

The proposed system reduces risk and increases level of safety when applied in hazardous environment such as the nuclear one.

The contribution and uniqueness of the presented study are represented in the development of a well-integrated HRI system that can tackle the four aforementioned circumstances in an effective and user-friendly way. High operator–robot reactivity is kept by using the direct control method, while a lot of cognitive stress is removed using elective/flapped autonomous mode to manipulate randomly localized different configuration objects. This necessitates building an effective deep learning algorithm (in comparison to well-known methods) to recognize objects in different conditions: illumination levels, shadows and different postures.

With the development of deep learning-based analytical techniques, increased research has focused on fatigue data analysis methods based on deep learning, which are gaining in popularity. However, the application of deep neural networks in the material science domain is mainly inhibited by data availability. In this paper, to overcome the difficulty of multifactor fatigue life prediction with small data sets,

A multiple neural network ensemble (MNNE) is used, and an MNNE with a general and flexible explicit function is developed to accurately quantify the complicated relationships hidden in multivariable data sets. Moreover, a variational autoencoder-based data generator is trained with small sample sets to expand the size of the training data set. A comparative study involving the proposed method and traditional models is performed. In addition, a filtering rule based on the R2 score is proposed and applied in the training process of the MNNE, and this approach has a beneficial effect on the prediction accuracy and generalization ability.

A comparative study involving the proposed method and traditional models is performed. The comparative experiment confirms that the use of hybrid data can improve the accuracy and generalization ability of the deep neural network and that the MNNE outperforms support vector machines, multilayer perceptron and deep neural network models based on the goodness of fit and robustness in the small sample case.

The experimental results imply that the proposed algorithm is a sophisticated and promising multivariate method for predicting the contact fatigue life of a coating when data availability is limited.

A data generated model based on variational autoencoder was used to make up lack of data. An MNNE method was proposed to apply in the small data case of fatigue life prediction.

Heavy equipment residual value forecasting is dynamic as it relies on the age, type, brand and model of the equipment, ranking condition, place of sale, operating hours and other macroeconomic gauges. The main objective of this study is to predict the residual value of the main types of heavy construction equipment. The residual value of heavy construction equipment is predicted via deep learning (DL) and machine learning (ML) approaches.

Based on deep and machine learning regression network integrated with data mining, random forest (RF), decision tree (DT), deep neural network (DNN) and linear regression (LR)-based modeling decision support models are developed. This research aims to forecast the residual value for different types of heavy construction equipment. A comprehensive investigation of publicly accessible auction data related to various types and categories of construction equipment was utilized to generate the model's training and testing datasets. In total, four performance metrics (i.e. the mean absolute error (MAE), mean squared error (MSE), the mean absolute percentage error (MAPE) and coefficient of determination (R2)) were used to measure and compare the developed algorithms' accuracy.

The developed algorithm's efficiency has been demonstrated by comparing the deep and machine learning predictions with real residual value. The accuracy of the results obtained by different proposed modeling techniques was comparable based on the performance evaluation metrics. DT shows the highest accuracy of 0.9111 versus RF with an accuracy of 0.8123, followed by DNN with an accuracy of 0.7755 and the linear regression with an accuracy of 0.5967.

The proposed novel model is designed as a supportive tool for construction project managers for equipment selling, purchasing, overhauling, repairing, disposing and replacing decisions.

Because of the mechanical properties of aluminium (Al), an accurate prediction of its properties has been challenging. Researchers are seeking reliable models for predicting the mechanical strength of Al alloys owing to the continuous emergence of new Al alloys and their applications. There has been widespread use of empirical and statistical models for the prediction of different mechanical properties of Al and Al alloy, such as linear and nonlinear regression. Nevertheless, the development of these models requires laborious experimental work, and they may not produce accurate results depending on the relationship between the Al properties, mix of other compositions and curing conditions.

Numerous machine learning (ML) models have been proposed as alternative approaches for predicting the strengths of Al and its alloys. The hardness of Al alloys has been predicted by implementing various ML algorithms, such as linear regression, ridge regression, lasso regression and artificial neural network (ANN). This investigation critically analysed and discussed the application and performance of models generated by linear regression, ridge regression, lasso regression and ANN algorithms using different mechanical properties as training parameters.

Considering the definition of the problem, linear regression has been found to be the most suitable algorithm in predicting the hardness values of AA7XXX alloys as the model generated by it best fits the data set.

The work presented in this paper is original and not submitted anywhere else.

The main goal of the physical education (PE) environment is that each individual trained should achieve self-fulfillment with the large group of students involved with their own efforts. Deep learning is applying transferrable knowledge in new situations to help the students master in tough circumstances. In PE training, injuries occur when working together as a team. Safety measures are taken immediately as an emergency response to reduce the potential risk in students by providing first aid. To provide safety measures for the injured student immediately, the environment is monitored in real-time using a GPS.

Theory of Humanities Education (ToHE) infers that it has less collection of theories and a wide range of applications than the state-of-the-art systems. ToHE allows students to think creatively and play a vital role in one’s health which is a critical aspect in PE. The ToHE theory focuses on two main concepts, i.e. by using a methodological approach to analyse and deep learning to solve the problem. PE motivates college students to follow a healthy and active lifestyle.

The proposed system is deployed in real time for monitoring the student’s performance and provides an emergency response with an accuracy rate of 90%.

The deep learning offers solutions to the injuries by using the deep convolutional neural network to provide interpretability of the consequence by training it with various injuries that occur in the playground and inappropriate use of sports equipment. A case study provided in this paper outlines an emergency response scenario to an injured student in sports training.

Recently, the convolutional neural network (ConvNet) has a wide application in the classification of motor imagery EEG signals. However, the low signal-to-noise electroencephalogram (EEG) signals are collected under the interference of noises. However, the conventional ConvNet model cannot directly solve this problem. This study aims to discuss the aforementioned issues.

To solve this problem, this paper adopted a novel residual shrinkage block (RSB) to construct the ConvNet model (RSBConvNet). During the feature extraction from EEG signals, the proposed RSBConvNet prevented the noise component in EEG signals, and improved the classification accuracy of motor imagery. In the construction of RSBConvNet, the author applied the soft thresholding strategy to prevent the non-related motor imagery features in EEG signals. The soft thresholding was inserted into the residual block (RB), and the suitable threshold for the current EEG signals distribution can be learned by minimizing the loss function. Therefore, during the feature extraction of motor imagery, the proposed RSBConvNet de-noised the EEG signals and improved the discriminative of classification features.

Comparative experiments and ablation studies were done on two public benchmark datasets. Compared with conventional ConvNet models, the proposed RSBConvNet model has obvious improvements in motor imagery classification accuracy and Kappa coefficient. Ablation studies have also shown the de-noised abilities of the RSBConvNet model. Moreover, different parameters and computational methods of the RSBConvNet model have been tested on the classification of motor imagery.

Based on the experimental results, the RSBConvNet constructed in this paper has an excellent recognition accuracy of MI-BCI, which can be used for further applications for the online MI-BCI.