Deep Convolutional Network Research Articles

Reconstruction of Fine-Spatial-Resolution FY-3D-Based Vegetation Indices to Achieve Farmland-Scale Winter Wheat Yield Estimation via Fusion with Sentinel-2 Data

The spatial resolution (250–1000 m) of the FY-3D MERSI is too coarse for agricultural monitoring at the farmland scale (20–30 m). To achieve the winter wheat yield (WWY) at the farmland scale, based on FY-3D, a method framework is developed in this work. The enhanced deep convolutional spatiotemporal fusion network (EDCSTFN) was used to perform a spatiotemporal fusion on the 10 day interval FY-3D and Sentinel-2 vegetation indices (VIs), which were compared with the enhanced spatial and temporal adaptive reflectance fusion model (ESTARFM). In addition, a BP neural network was built to calculate the farmland-scale WWY based on the fused VIs, and the Aqua MODIS gross primary productivity product was used as ancillary data for WWY estimation. The results reveal that both the EDCSTFN and ESTARFM achieve satisfactory precision in the fusion of the Sentinel-2 and FY-3D VIs; however, when the period of spatiotemporal data fusion is relatively long, the EDCSTFN can achieve greater precision than ESTARFM. Finally, the WWY estimation results based on the fused VIs show remarkable correlations with the WWY data at the county scale and provide abundant spatial distribution details about the WWY, displaying great potential for accurate farmland-scale WWY estimations based on reconstructed fine-spatial-temporal-resolution FY-3D data.

Reliable Augmentation and Precise Identification of EPG Waveforms Based on Multi-Criteria DCGAN

The electrical penetration graph (EPG) technique is of great significance in elucidating the mechanisms of virus transmission by piercing-sucking insects and crop resistance to these insects. The traditional method of manually processing EPG signals encounters the drawbacks of inefficiency and subjectivity. This study investigated the data augmentation and automatic identification of various EPG signals, including A, B, C, PD, E1, E2, and G, which correspond to distinct behaviors exhibited by the Asian citrus psyllid. Specifically, a data augmentation method based on an improved deep convolutional generative adversarial network (DCGAN) was proposed to address the challenge of insufficient E1 waveforms. A multi-criteria evaluation framework was constructed, leveraging maximum mean discrepancy (MMD) to evaluate the similarity between the generated and real data, and singular value decomposition (SVD) was incorporated to optimize the training iterations of DCGAN and ensure data diversity. Four models, convolutional neural network (CNN), K-nearest neighbors (KNN), decision tree (DT), and support vector machine (SVM), were established based on DCGAN to classify the EPG waveforms. The results showed that the parameter-optimized DCGAN strategy significantly improved the model accuracies by 5.8%, 6.9%, 7.1%, and 7.9% for DT, SVM, KNN, and CNN, respectively. Notably, DCGAN-CNN effectively addressed the skewed distribution of EPG waveforms, achieving an optimal classification accuracy of 94.13%. The multi-criteria optimized DCGAN-CNN model proposed in this study enables reliable augmentation and precise automatic identification of EPG waveforms, holding significant practical implications for understanding psyllid behavior and controlling citrus huanglongbing.

An Adaptive Framework for Traffic Congestion Prediction using Deep Learning

Aim and background: Congestion on China's roads has worsened in recent years due to the country's rapid economic development, rising urban population, rising private car ownership, inequitable traffic flow distribution, and growing local congestion. As cities expand, traffic congestion has become an unavoidable nuisance that endangers the safety and progress of its residents. Improving the utilization rate of municipal transportation facilities and relieving traffic congestion depend on a thorough and accurate identification of the current state of road traffic and necessitate anticipating road congestion in the city. Methodology: In this research, we suggest using a deep spatial and temporal graph convolutional network (DSGCN) to forecast the current state of traffic congestion. To begin, we grid out the transportation system to create individual regions for analysis. In this work, we abstract the grid region centers as nodes, and we use an adjacency matrix to signify the dynamic correlations between the nodes. Results and Discussion: The spatial correlation between regions is then captured utilizing a Graph Convolutional-Neural-Network (GCNN), while the temporal correlation is captured using a two-layer long and short-term feature model (DSTM). Conclusion: Finally, testing on real PeMS datasets shows that the DSGCN has superior performance than other baseline models and provides more accurate traffic congestion prediction.

SkipNode: On Alleviating Performance Degradation for Deep Graph Convolutional Networks

SkipNode: On Alleviating Performance Degradation for Deep Graph Convolutional Networks

LW-DCGAN: a lightweight deep convolutional generative adversarial network for enhancing occluded face recognition

LW-DCGAN: a lightweight deep convolutional generative adversarial network for enhancing occluded face recognition

HRTBDA: a network for post-disaster building damage assessment based on remote sensing images

ABSTRACT Efficient building damage assessment after disasters is vital for emergency response and loss evaluation, but the task is complicated by diverse building structures and complex environments. Traditional methods using Convolutional Neural Networks (CNNs) struggle to capture global contextual features, limiting damage categorization accuracy. To address this, we introduce the High-Resolution Transformer Architecture for Building Damage Assessment (HRTBDA), which enhances multi-scale feature extraction. A Cross-Attention-Based Spatial Fusion (CSF) module is proposed to utilize the attention mechanism, improving the model’s ability to identify detailed associations in damaged buildings. Additionally, we propose a deep convolution network matching optimization strategy that integrates a multilayer perceptron and expands the receptive field, enhancing global feature perception. HRTBDA’s performance was evaluated on two public datasets and compared with five recent frameworks. The model achieved an F1-score of 86.0% in building localization and 78.4% in damage assessment, with a 4.8% improvement in detecting minor damages. These results demonstrate HRTBDA’s potential for improving building damage assessment and highlight its significant advancements over existing methods.

Improving losses & accuracy through design of deep convolutional generative adversarial network (DCGAN) for plant disease detection tasks

Improving losses & accuracy through design of deep convolutional generative adversarial network (DCGAN) for plant disease detection tasks

Improving Recognition of Road Users via Doppler Radar Data and Deep Learning Convolutional Networks

This research presents findings from laboratory experiments on a novel method for identifying and differentiating objects using radar signatures and a specialized convolutional neural network architecture. Previously introduced by the authors, this method has been validated through real-world measurements in an urban environment with a 24 GHz frequency-modulated continuous-wave radar. This study describes how radar signatures, generated in the MATLAB (R2023b) environment from I and Q signals captured by the uRAD USB v1.2 radar, were processed. A database of radar signatures for pedestrians, cyclists, and vehicles was created, and a tailored convolutional neural network was trained. The developed solution achieves an accuracy of over 95% in distinguishing between various objects. The simulation results and successful tests support the application of this system across various sectors. The key innovations include distinguishing multiple objects from a single radar signature, a custom architecture for the convolutional neural network, and an application that processes radar data to produce near-real-time recognition results.

Physics-informed deep convolutional network for combined sea ice concentration and velocity prediction

Accurate prediction of short-term fluctuations in Arctic sea ice is important for safe use of Arctic shipping routes. While deep-learning models can improve the accuracy of sea-ice predictions, the accuracy and predictability of short-term sea-ice conditions are limited in most purely data-driven deep-learning models that consider a single aspect of sea ice, without considering the physical variation law between several sea ice factors. We propose dual-task prediction models for sea ice concentration (SIC) and sea ice velocity (SIV), and incorporate a loss function that simultaneously addresses dynamic constraints of SIC and SIV into each model, based on dynamics terms in the sea ice control equation. Comparative experiments are performed to determine which model structure best performs at predicting SIC and SIV. We report a dual-task branching structure to be more suitable for predicting SIC and SIV, and a post-decoder branch network structure to best predict SIC and SIV. Adding a sea ice dynamics equation to the loss function improves the model fit with dynamics law, improving the predictability and prediction accuracy of SIC and SIV.

An efficient attention-based hybridized deep learning network with deep RBM features for customer behavior prediction in digital marketing

PurposeAn efficient customer behavior prediction model is designed using deep learning techniques. The necessary data used for the implementation are taken from standard datasets and presented to perform subsequent tasks. Here, deep restricted Boltzmann machines (RBM) features are retrieved from the input images. Further, the extracted deep RBM features are presented to the customer behavior prediction phase. Here, the attention-based hybrid deep learning (A-HDL) technique is designed based on the incorporation of a dilated deep temporal convolutional network (dilated-DTCN) and a weighted recurrent neural network (weighted RNN). Moreover, the weights in RNN are tuned using a modernized random parameter-based cheetah optimizer (MRPCO). Further, various experiments were performed on the implemented framework, and it secured an enhanced customer behavior prediction rate than the conventional models.Design/methodology/approachA novel hybrid deep network-based customer behavior prediction model was developed to predict the behavior of the customer so the companies yield more income by advertising their products based on the predicted results.FindingsWhen considering the first dataset, the designed customer behavior prediction mechanism produced 94% accuracy, which is higher than the conventional techniques such as long short-term memory (LSTM), DTCN, RNN and A-HDL with 88%, 87%, 89% and 93%.Originality/valueThe precision and the accuracy of the developed MRPCO-A-HDL-based customer behavior prediction model progressed than the conventional techniques and algorithms.

Design and Research on Identification of Typical Tea Plant Diseases Using Small Sample Learning

Tea plants are susceptible to diseases during their growth. These diseases seriously affect the yield and quality of tea. The effective prevention and control of diseases requires accurate identification of diseases. With the development of artificial intelligence and computer vision, automatic recognition of plant diseases using image features has become feasible. As the support vector machine (SVM) is suitable for high dimension, high noise, and small sample learning, this paper uses the support vector machine learning method to realize the segmentation of disease spots of diseased tea plants. An improved Conditional Deep Convolutional Generation Adversarial Network with Gradient Penalty (C-DCGAN-GP) was used to expand the segmentation of tea plant spots. Finally, the Visual Geometry Group 16 (VGG16) deep learning classification network was trained by the expanded tea lesion images to realize tea disease recognition.

A comprehensive investigation on the performance of reconstruction of noncircular fiber-representative volume elements in unidirectional composites using diffusion generative models

This study employs diffusion generative models to reconstruct random representative volume elements (RVEs) of unidirectional composites with noncircular fibers. Microscope images of these composites were prepared and trained with denoising diffusion probabilistic model (DDPM), denoising diffusion implicit model (DDIM), progressive distillation diffusion model (PDDM), and deep convolutional generative adversarial network (DCGAN). Hyperparameter tuning was performed for both DDPM and PDDM, and the generated RVE images were evaluated using the two-point correlation function (TPCF), Fréchet Inception Distance (FID), and computational cost. Furthermore, finite element (FE) models were generated using these images, and FE simulations were conducted considering interfacial debonding behavior. The resulting stress and strain curves from these simulations were compared. The results show that DDPM demonstrated the best performance in final image quality, while PDDM maintained stable performance from the early stages of training. Additionally, both models exhibited excellent agreement with the original images, indicating high quality, diversity, and resemblance.

A deep learning‐based channel estimation and joint adaptive hybrid beamforming approach for mm‐wave massive MIMO with group‐of‐subarrays

SummaryMillimeter‐wave massive multiple‐input‐multiple‐output (MIMO) is a wireless communication that enables communication at extremely high data rates. It is widely used to expand communication bandwidth to improve spectral and energy efficiency. The mm‐wave system is usually based on the radio frequency (RF) chain. Beamforming is used to minimize the interface signal and maximize each user's received signal power. The hybrid beamforming method uses channel estimation and beam generation techniques that improve overall performance and save energy. There are various limitations and disadvantages arising from different techniques of the MIMO system. Hence, the proposed model is introduced to enhance the performance and maximize the spectral and energy efficiency. Initially, the optimal RF chain selection and channel optimization are done by the differential evolution firefly‐assisted optimized channel compression‐reconstruction (DEF_OCCR) network. In this research, the RF chain selection process is done using hybrid differential evolution firefly optimization. A robust autoencoder‐driven deep learning model is proposed for channel estimation after RF chain selection. Finally, the deep learning convolutional joint adaptive hybrid beamforming network (DConvHB) is used for beamforming to maximize spectral efficiency. The overall performance of this work is analyzed using several existing models to describe its superiority. The average system throughput is 12.85 bits, and the spectral efficiency of the proposed model is 5 bits for 10 dB and 32.18 bits for 13 dB. The proposed energy efficiency model is 1.73 bits and 4.96 bits for 10 dB.

An improved deep convolutional generative adversarial network for quantification of catechins in fermented black tea

The rapid and non-destructive quantification of catechins in fermented black tea is crucial for evaluating the quality of black tea. The combination of hyperspectral imaging and chemometrics has been applied for quantitative detection, but its performance is usually constrained by the limited dataset size. Targeted at the challenge of insufficient samples in regression analysis of catechins, this study proposes an improved deep convolutional generative adversarial network with labeling module, named as DCGAN-L for hyperspectral data augmentation. The DCGAN-L consists of the spectral and label generating modules. First the synthetic spectra were generated, and an indicator was proposed to evaluate their quality. Then, the corresponding label values were generated, including epicatechin gallate (ECG), epicatechin (EGC), catechin (C), and total catechin (CC). For label generating, the Euclidean distances between the synthetic spectrum and all true spectra were measured, followed by allocating weights for calculating the label values based on these distances. Subsequently, the training dataset was augmented with the generated synthetic data. The effect of data augmentation was finally evaluated based on two regression models of random forest (RF) and broad learning system (BLS) for the quantification of catechins. Compared with the results before data augmentation, the average R2 of RF and BLS models increased by 0.044 and 0.164, respectively. The proposed DCGAN-L model allows for the rapid, non-destructive quantification of catechins in black tea in the case of limited sample size.

DEEP Q-NETWORK (DQN) PARTIALOCCLUSION SEGMENTATION AND BACKTRACKING SEARCH OPTIMIZATION ALGORITHM (BSOA) WITH OPTICAL FLOW RECONSTRUCTION FOR FACIAL EXPRESSION EMOTION RECOGNITION

Video facial expression recognition (FER) has garnered a lot of attention recently and is helpful for several applications. Although many algorithms demonstrate impressive performance in a controlled environment without occlusion, identification in the presence of partial facial occlusion remains a challenging issue. Solutions based on reconstructing the obscured area of the face have been suggested as a way to deal with occlusions. These options mostly rely on the face’s shape or texture. Nonetheless, the resemblance in facial expressions among individuals appears to be a valuable advantage for the reconstruction. For semantic segmentation based on occlusions, Reinforcement Learning (RL) is introduced as the initial stage. From a pool of unlabeled data, an agent learns a policy to choose a subset of tiny informative image patches to be tagged instead of full images. In the second stage, a trained Backtracking Search Algorithm (BSA) is used to rebuild optical flows that have been distorted by the occlusion. On obtaining optical flows estimated from occluded facial frames, AEs restore optical flows of occluded regions. These recovered optical flows become inputs to anticipate classes f expressions. Optical flux reconstructions then classify stages. This study evaluates classification model’s performances for face expression identification based on Very Deep Convolution Networks (VGGNet). Furthermore, it produces more accurate confusion matrices and proposes approaches for the KMU-FED and CK+ databases, respectively. The results are evaluated using metrics including recall, f-measure, accuracy, and precision.

Synthetic Image Generation Using Deep Learning: A Systematic Literature Review

ABSTRACTThe advent of deep neural networks and improved computational power have brought a revolutionary transformation in the fields of computer vision and image processing. Within the realm of computer vision, there has been a significant interest in the area of synthetic image generation, which is a creative side of AI. Many researchers have introduced innovative methods to identify deep neural network‐based architectures involved in image generation via different modes of input, like text, scene graph layouts and so forth to generate synthetic images. Computer‐generated images have been found to contribute a lot to the training of different machine and deep‐learning models. Nonetheless, we have observed an immediate need for a comprehensive and systematic literature review that encompasses a summary and critical evaluation of current primary studies' approaches toward image generation. To address this, we carried out a systematic literature review on synthetic image generation approaches published from 2018 to February 2023. Moreover, we have conducted a systematic review of various datasets, approaches to image generation, performance metrics for existing methods, and a brief experimental comparison of DCGAN (deep convolutional generative adversarial network) and cGAN (conditional generative adversarial network) in the context of image generation. Additionally, we have identified applications related to image generation models with critical evaluation of the primary studies on the subject matter. Finally, we present some future research directions to further contribute to the field of image generation using deep neural networks.

Identification method for micro–nano defects in Si3N4 ceramic bearing rollers

In response to the surface micro–nano defects of small Si3N4 ceramic bearing rollers during use, as well as the poor generalization ability and low accuracy of the object detection model trained on the micro and nano defect dataset, an enhanced recognition algorithm based on deep convolutional generative adversarial networks is proposed. Due to the limited size of the dataset, the DCGAN model is reconstructed to effectively expand the micro–nano defect dataset. In addition, noise generalization is applied to stabilize DCGAN model training, creating a low-dimensional manifold distribution to ensure significant overlap between the data and the original dataset, and activating Jensen–Shannon (JS) divergence for stable training. To verify the effectiveness of the enhanced dataset, synthetic micro–nano defects are used to improve the YOLO-v4-tiny object detection model. By comparing t-distributed stochastic neighbor embedding (t-SNE) and feature vectors, it can be found that the images generated by the optimized DCGAN have higher grayscale feature diversity and better visual consistency. After generations of enhancements, the micro–nano defect detection speed has reached 226FPS, and the accuracy has reached 97.41%.

Advancing Photorealism Enhancement Using Convolutional Neural Networks (CNNs)

Photorealism has been the defining goal of computer graphics for half a century. A look at even the most sophisticated real-time games will quickly reveal that photorealism has not been achieved. An ineffable difference in the appearance of simulation and reality remains. In recent years, a complementary set of techniques has been developed in computer vision and machine learning. These techniques, based on deep learning, convolutional networks, and adversarial training, bypass physical modeling of geometric layout, material appearance, and light transport. Instead, images are synthesized by convolutional networks trained on large datasets. These techniques have been used to synthesize representative images from a given domain to convert semantic label maps to photographic images and to attempt to bridge the appearance gap between synthetic and real images. The images are enhanced by a convolutional network that leverages intermediate representations produced by conventional rendering pipelines. The network is trained via a novel adversarial objective, which provides strong supervision at multiple perceptual levels.

Envelope spectrum neural network with adaptive domain weight harmonization for intelligent bearing fault diagnosis under cross-machine scenarios

Accurate bearing fault diagnosis technology is highly important for ensuring the safe operation of mechanical equipment. Fault diagnosis methods can be roughly divided into signal processing-based methods (SPM) and data-driven methods (DDMs), which rely on physical knowledge and data knowledge, respectively. However, SPM cannot adapt to large data samples and dynamic parameter variations. DDMs did not consider physical representation consistency. To address the above issues, an envelope spectrum neural network (ESNN) is proposed for cross-machine bearing fault diagnosis. First, a deep transfer convolution network (DTCN) is constructed as the basic diagnostic framework. Second, a knowledge-to-model conversion strategy (KMC) is built to create ESNN, which utilizes equivalent convolution functions to construct the first two layers of DTCN, guiding the model to learn physical knowledge. Subsequently, an adaptive domain weight harmonization (ADWH) mechanism is proposed that can dynamically combine marginal distributions and joint distributions and automatically learn hidden features contained in physical and data knowledge, thus alleviating domain shift issues. Experimental evaluations are conducted using fault datasets from three different machines. The results show that, compared to models without knowledge guidance, ESNN achieves a 6.4% improvement in diagnostic accuracy. Compared to many advanced models, ESNN can achieve a maximum diagnostic accuracy improvement of 35.94%. The code of the ESNN can be found at https://github.com/John-520/ESNN.

-AI-assisted diagnostic potential of CT in bone oncology and its impact on clinical decision-making for intensive care

ObjectiveThis study evaluates the AI-assisted diagnostic potential of computed tomography (CT) for bone cancer and its influence on patient care during the pre- and post-treatment phases. It compares patient management approaches based on CT severity levels and identifies distinct CT phenotypes linked to disease severity. MethodologyWe retrospectively examined 50 patients diagnosed with bone cancer between December 2022 and June 2023. The CT scans were analyzed according to the Radiological Society of North America (RSNA) guidelines. This study was performed using the deep convolutional neutral network (DCNN) model to assist doctors in diagnosing bone tumors through CT scanning. Patients’ management approaches were compared based on the severity levels indicated by CT scans. ResultsFifty patients participated in this study, with a median age of 67.2 years, ranging from 32 to 89 years. Of them, 38 % were female and 62 % were male. In 2022, 19 individuals (13 males and 6 females, ages 32 to 84) were assessed, with a mean age of 59.9 years. In 2023, 31 individuals, aged 54 to 89 with a mean age of 71.6 years, were assessed; among them were 18 men and 13 women. SPECT scans revealed the following key diagnostic features: 85.9 % of patients exhibited bone lesions with ground-glass opacities, 88 % had multipolar involvement, 92.8 % had bilateral involvement, and 92.8 % showed peripheral involvement. The severity scores based on CT scans were significantly higher in patients requiring intensive care, with scores above 14 being more common in this group. ConclusionDistinct CT findings during the AI-assisted diagnosis and treatment of bone cancer provided prompt and sensitive examination capabilities. Notably, two CT phenotypes emerged, associated with large consolidation patterns and high severity scores, offering crucial insights into disease severity and aiding in clinical decision-making for intensive care requirements. The study underscores the importance of CT in the effective monitoring and management of bone cancer pre- and post-treatment.