Sentinel-2 Sharpening Using a Single Unsupervised Convolutional Neural Network With MTF-Based Degradation Model

The world has recently seen unprecedented growth of social networks that have enabled internet users to share and communicate their views in various areas of life. Analysis of opinion is one type of text mining that aims to discover human views and emotions about any particular event, or topic. In the experiment of this research, we perform document level sentiment analysis in the form of positive and negative sentiments using both the traditional models and recent deep learning models. Three traditional methods (i.e. Naive Bayes (NB), Logistic Regression (LR), and Voting Ensemble) and three deep learning layout (i.e. Single Convolutional Neural Network (CNN), Single CNN + FastText and Ensemble(2-Layer CNN + Gated Recurrent Unit (GRU))) have been employed for two different datasets to evaluate the opinion classification. In our experiment, accuracy, f1-score, and RMSE have been considered as evaluation parameters and the optimistic results show that for airline review dataset, Single CNN + FastText comes out with the highest performance whereas for hotel review dataset, 2-L CNN + GRU comes out with the highest performance.

Vision-based automatic detection and tracking of on-road vehicles is a demanding research area in intelligent transportation systems (ITS). Several investigation reports are available in the literature in this research field using various machine learning (ML) techniques. This article proposes a method for on-road vehicle detection and tracking in varying weather conditions by modifying the structure of you only look once (YOLO), a recently developed deep learning model in computer vision for detecting objects. ML-based systems repurpose classifiers to detect the vehicles, whereas in YOLO model, a single convolutional neural network (CNN) predicts the detecting bounding boxes and the class probabilities for those boxes together in one evaluation from the full images. As the entire detection is performed using a single network, the detection process can be viewed as an end-to-end system. In this work, 16 convolutional layers of YOLO have been used along with two fully connected layers at the end. The vehicles have been detected in varying weather conditions. Two different public data sets, namely CDNet 2014 and LISA 2010 data sets have been used to evaluate the system performance, and it outperforms state-of-the-art results.

The study intended to explore the effect of different neural network algorithms in the electrocardiogram (ECG) classification of patients with congenital heart disease (CHD). Based on the single convolutional neural network (CNN) ECG algorithm and the recurrent neural network (RNN) ECG algorithm, a multimodal neural network (MNN) ECG algorithm was constructed utilizing the MIT-BIH database as training set and test set. Furthermore, the MNN ECG algorithm was optimized to establish an improved MNN (IMNN) algorithm, which was applied to the diagnosis of CHD patients. The CHD patients admitted between August 2016 and August 2019 were selected for analysis to compare the classification effect and accuracy rate of IMNN, MNN, CNN ECG, and RNN ECG algorithms. It was found that the RNN ECG algorithm had higher classification sensitivity and true positive rate in terms of normal or bundle (NB) branch block beat, supraventricular abnormal (SA) rhythm, abnormal ventricular (AV) beat, and fusion beat (FB) than the CNN ECG algorithm ( P < 0.05 ), and the classification sensitivity and true positive rate of IMNN algorithm in the four aspects were significantly higher than those of MNN algorithm ( P < 0.05 ). The classification accuracy of CNN ECG algorithm and RNN ECG algorithm was above 98%, while that of MNN algorithm and IMNN algorithm was better than that of CNN ECG algorithm and RNN ECG algorithm, and the accuracy rate can reach 98.5% or more. Moreover, the accuracy rate of the IMNN algorithm can reach more than 98%. In conclusion, IMNN not only has a good classification ability in the simulated environment but also performs well in the actual environment, which is worthy of clinical promotion.

In the domain of chemometrics, multiblock data analysis is widely performed for exploring or fusing data from multiple sources. Commonly used methods for multiblock predictive analysis are the extensions of latent space modelling approaches. However, recently, deep learning (DL) approaches such as convolutional neural networks (CNNs) have outperformed the single block traditional latent space modelling chemometric approaches such as partial least-square (PLS) regression. The CNNs based DL modelling can also be performed to simultaneously deal with the multiblock data but was never explored until this study. Hence, this study for the first time presents the concept of parallel input CNNs based DL modelling for multiblock predictive chemometric analysis. The parallel input CNNs based DL modelling utilizes individual convolutional layers for each data block to extract key features that are later combined and passed to a regression module composed of fully connected layers. The method was tested on a real visible and near-infrared (Vis-NIR) large data set related to dry matter prediction in mango fruit. To have the multiblock data, the visible (Vis) and near-infrared (NIR) parts were treated as two separate blocks. The performance of the parallel input CNN was compared with the traditional single block CNNs based DL modelling, as well as with a commonly used multiblock chemometric approach called sequentially orthogonalized partial least-square (SO-PLS) regression. The results showed that the proposed parallel input CNNs based deep multiblock analysis outperformed the single block CNNs based DL modelling and the SO-PLS regression analysis. The root means squared errors of prediction obtained with deep multiblock analysis was 0.818%, relatively lower by 4 and 20% than single block CNNs and SO-PLS regression, respectively. Furthermore, the deep multiblock approach attained ∼3% lower RMSE compared to the best known on the mango data set used for this study. The deep multiblock analysis approach based on parallel input CNNs could be considered as a useful tool for fusing data from multiple sources.

As the number of social media comments available online grows, the spread of hate speech has grown gradually. When someone uses hate speech as a weapon to injure, degrade, and humiliate others, their freedom, dignity, and personhood can be jeopardized. Deep neural network-based hate speech detection models, such as the conventional single channel convolutional neural network (SC-CNN), have recently demonstrated promising results. The success of the models, however, is dependent on the type of language they are trained on and the training data size. Even with a small amount of training data, the model's performance can be improved by using a multichannel convolutional neural network (MC-CNN) model. The study assesses and compares the performance of a multichannel convolutional neural network model to single channel convolutional neural network models using a support vector machine (SVM) as a baseline. The models' F1 score values are computed, and promising results are obtained. The MC-CNN model outperforms the SC-CNN models in all three hate speech datasets. The study's findings indicate that the proposed MC-CNN model could be used as a deep learning-based alternative for hate speech detection.

The estimation of crowd density is a crucial area of study in computer vision because of its extensive use in intelligence collection, urban planning, and surveillance. The crowd density prediction that comes from a thorough analysis considers several factors, including inter-blocking in dense crowds, background elements, and individual appearance similarity. To monitor populous areas and avoid congestion, we are interested in using machine learning for crowd control in this research. We suggest using a Single Convolutional Neural Network with Three Layers (S-CNN3) model to estimate the crowd size and count the number of individuals in a scene. Next, a comparison study for density counting determines how well the suggested model performs in comparison to switched convolutional neural networks (SCNN) and convolutional neural networks with four layers (single CNN4). This study makes use of the Shanghai Tech dataset, which is regarded as the biggest database for crowd counting. With an average test accuracy of 99.88% and an average validation loss of 0.02 for crowd density estimate, the suggested model demonstrates remarkable efficacy and efficiency. These outcomes outperform the state-of-the-art models already in use.

Deep neural networks, albeit their great success on feature learning in various computer vision tasks, are usually considered as impractical for online visual tracking, because they require very long training time and a large number of training samples. In this paper, we present an efficient and very robust tracking algorithm using a single convolutional neural network (CNN) for learning effective feature representations of the target object in a purely online manner. Our contributions are multifold. First, we introduce a novel truncated structural loss function that maintains as many training samples as possible and reduces the risk of tracking error accumulation. Second, we enhance the ordinary stochastic gradient descent approach in CNN training with a robust sample selection mechanism. The sampling mechanism randomly generates positive and negative samples from different temporal distributions, which are generated by taking the temporal relations and label noise into account. Finally, a lazy yet effective updating scheme is designed for CNN training. Equipped with this novel updating algorithm, the CNN model is robust to some long-existing difficulties in visual tracking, such as occlusion or incorrect detections, without loss of the effective adaption for significant appearance changes. In the experiment, our CNN tracker outperforms all compared state-of-the-art methods on two recently proposed benchmarks, which in total involve over 60 video sequences. The remarkable performance improvement over the existing trackers illustrates the superiority of the feature representations, which are learned purely online via the proposed deep learning framework.

The three seismic image processing tasks of fault detection, structure-oriented smoothing with edge preserving, and slope estimation are related to each other and all involve analyzing seismic structural features. In conventional processing schemes, however, these three tasks are often independently preformed by different algorithms and challenges remain in each of them. We propose to simultaneously perform all the three tasks by using a single convolutional neural network. To train the network, we automatically create thousands of 3D synthetic noisy seismic images and the corresponding ground truth of fault images, clean seismic images, and seismic normal vectors. Although trained with only synthetic datasets, the network automatically learns to accurately perform all the three image processing tasks in a general seismic image. Multiple field examples show that the network is significantly superior to the conventional methods in all the three tasks of computing a more accurate and sharper fault detection, a smoothed seismic volume with better enhanced structures and structural edges, and more accurate reflection slopes. Presentation Date: Monday, September 16, 2019 Session Start Time: 1:50 PM Presentation Time: 4:45 PM Location: 221D Presentation Type: Oral

Crowd density estimation is an important topic in computer vision due to its widespread applications in surveillance, urban planning, and intelligence gathering. Resulting from extensive analysis, crowd density estimation reflects many aspects such as similarity of appearance between people, background components, and inter-blocking in intense crowds. In this paper, we are interested to apply machine learning for crowd management in order to monitor populated area and prevent congestion situations. We propose a Single-Convolutional Neural Network with Three Layers (S-CNN3) model to count the number of people in a scene and conclude about the crowd estimation. Then, a comparative study for density counting establishes the performance of the proposed model against the convolutional neural networks with four layers (single-CNN4) and Switched Convolutional neural networks (SCNN). ShanghaiTech dataset, considered as the largest data base for crowd counting, is used in this work. The proposed model proves high effectiveness and efficiency for crowd density estimation with 99.88% of average test accuracy and 0.02 of average validation loss. These results achieve better performance than the existing state-of-the-art models.

Ionospheric delay poses significant challenges for satellite‐based communication and navigation applications. Predicting the global ionosphere variation can help to capture the vertical total electron content (VTEC) in advance, which improves the signal stability of communication and navigation, and supports these space‐related technologies. In existing studies, the ConvLSTM‐based method has achieved impressive results in global ionosphere map (GIM) prediction and gained increasing attention. However, the ConvLSTM model can only consider local features in GIM prediction since the single layer Convolutional Neural Network structure in ConvLSTM can not capture the global features. Inspired by the ResNet and the SimVP models, this paper proposes a new TEC prediction method based on a large receptive field model named large receptive field model with residual structure (LARRES). This method achieves a large receptive field effect through a deep global spatio‐temporal convolutional network, which allows to capture the global TEC features during the prediction process. The prediction performance of the LARRES model was evaluated with 8 years GIM data in both high and low solar activity years. The experimental results show that the proposed model outperforms the C1PG model from Center for Orbit Determination in Europe and two ConvLSTM‐based models. Compared to the C1PG model, the LARRES method can reduce RMS error by 28.7% and 10.4% during high and low solar activity years, respectively. The LARRES method can further reduce the global TEC prediction error by 0.35 TECU and 0.08 TECU subject to the ConvLSTM‐based models during high and low solar activity years respectively.

Although many machine learning methods have been widely used to predict PM2.5 concentrations, these single or hybrid methods still have some shortcomings. This study integrated the advantages of convolutional neural network (CNN) feature extraction and the regression ability of random forest (RF) to propose a novel CNN-RF ensemble framework for PM2.5 concentration modeling. The observational data from 13 monitoring stations in Kaohsiung in 2021 were selected for model training and testing. First, CNN was implemented to extract key meteorological and pollution data. Subsequently, the RF algorithm was employed to train the model with five input factors, namely the extracted features from the CNN and spatiotemporal factors, including the day of the year, the hour of the day, latitude, and longitude. Independent observations from two stations were used to evaluate the models. The findings demonstrated that the proposed CNN-RF model had better modeling capability compared with the independent CNN and RF models: the average improvements in root mean square error (RMSE) and mean absolute error (MAE) ranged from 8.10% to 11.11%, respectively. In addition, the proposed CNN-RF hybrid model has fewer excess residuals at thresholds of 10 μg/m3, 20 μg/m3, and 30 μg/m3. The results revealed that the proposed CNN-RF ensemble framework is a stable, reliable, and accurate method that can generate superior results compared with the single CNN and RF methods. The proposed method could be a valuable reference for readers and may inspire researchers to develop even more effective methods for air pollution modeling. This research has important implications for air pollution research, data analysis, model estimation, and machine learning.

Image denoising, a classic ill-posed problem, aims to recover a latent image from a noisy measurement. Over the past few decades, a considerable number of denoising methods have been studied extensively. Among these methods, supervised deep convolutional networks have garnered increasing attention, and their superior performance is attributed to their capability to learn realistic image priors from a large amount of paired noisy and clean images. However, if the image to be denoised is significantly different from the training images, it could lead to inferior results, and the networks may even produce hallucinations by using inappropriate image priors to handle an unseen noisy image. Recently, deep image prior (DIP) was proposed, and it overcame this drawback to some extent. The structure of the DIP generator network is capable of capturing the low-level statistics of a natural image using an unsupervised method with no training images other than the image itself. Compared with a supervised denoising model, the unsupervised DIP is more flexible when processing image content that must be denoised. Nevertheless, the denoising performance of DIP is usually inferior to the current supervised learning-based methods using deep convolutional networks, and it is susceptible to the over-fitting problem. To solve these problems, we propose a novel deep generative network with multiple target images and an adaptive termination condition. Specifically, we utilized mainstream denoising methods to generate two clear target images to be used with the original noisy image, enabling better guidance during the convergence process and improving the convergence speed. Moreover, we adopted the noise level estimation (NLE) technique to set a more reasonable adaptive termination condition, which can effectively solve the problem of over-fitting. Extensive experiments demonstrated that, according to the denoising results, the proposed approach significantly outperforms the original DIP method in tests on different databases. Specifically, the average peak signal-to-noise ratio (PSNR) performance of our proposed method on four databases at different noise levels is increased by 1.90 to 4.86 dB compared to the original DIP method. Moreover, our method achieves superior performance against state-of-the-art methods in terms of popular metrics, which include the structural similarity index (SSIM) and feature similarity index measurement (FSIM). Thus, the proposed method lays a good foundation for subsequent image processing tasks, such as target detection and super-resolution.

Based on the urgent need for agricultural intelligence in various northern countries, this study develops an algorithm for crop pest and disease identification based on the basic theory of convolutional neural networks (CNN). The existing related algorithms are basically based on platforms that can carry multiple cameras to work to (e.g., tractors or combine harvesters). Such algorithms are not applicable to cultivation scenarios such as terraces. Therefore, in this study, unlike other designs based on multiple nets, the learning model of this learning uses only a single net for the train dataset. The main research objective of this learning is to achieve satisfactory result using a simplified CNN model design. The main approach of this learning model is to use the convolutional layer in the model to learn the crop pest and disease features in the given images, then use the pooling layer to reduce the number of parameters, and finally use the Fully connected layer to classify them into two categories: diseased and healthy. After learning through the train dataset with a sample size of 7000, the test dataset was imported into the completed learned model for testing, and the result was an accuracy of 94.5% and a loss rate of 0.1385.

How to detect the collected P300 EEG signals more accurately is the core part of brain-computer interface technology. Aiming at the collected P300 EEG signals of the classical oddball paradigm, this paper proposed a fusion model based on CNN and RNN by using deep learning technology. This model adopts gated neural network GRU and combines attention mechanism to make the model better extract the time characteristics of EEG signals. At the same time, the model fuses the convolutional neural network to capture the spatial characteristics of EEG signals. The feature map extracted from the independent network is sent to the fully connected layer after splicing to obtain the classification results. Compared with some single convolutional neural networks, the proposed fusion model can better capture the time domain characteristics of signals. In terms of classification accuracy, the model proposed in this paper once reached 97.5%.

We present a deep-learning package named HiCNN2 to learn the mapping between low-resolution and high-resolution Hi-C (a technique for capturing genome-wide chromatin interactions) data, which can enhance the resolution of Hi-C interaction matrices. The HiCNN2 package includes three methods each with a different deep learning architecture: HiCNN2-1 is based on one single convolutional neural network (ConvNet); HiCNN2-2 consists of an ensemble of two different ConvNets; and HiCNN2-3 is an ensemble of three different ConvNets. Our evaluation results indicate that HiCNN2-enhanced high-resolution Hi-C data achieve smaller mean squared error and higher Pearson’s correlation coefficients with experimental high-resolution Hi-C data compared with existing methods HiCPlus and HiCNN. Moreover, all of the three HiCNN2 methods can recover more significant interactions detected by Fit-Hi-C compared to HiCPlus and HiCNN. Based on our evaluation results, we would recommend using HiCNN2-1 and HiCNN2-3 if recovering more significant interactions from Hi-C data is of interest, and HiCNN2-2 and HiCNN if the goal is to achieve higher reproducibility scores between the enhanced Hi-C matrix and the real high-resolution Hi-C matrix.