Network Training Research Articles

Adaptive decision-making with deep Q-network for heterogeneous unmanned aerial vehicle swarms in dynamic environments

With the continuous advancement of Unmanned Aerial Vehicle (UAV) swarm technology, the field of adaptive decision-making for heterogeneous UAV swarms across a spectrum of tasks is rapidly emerging as a focal point of research. This domain holds significant importance for enhancing the operational efficiency and responsiveness of UAV swarms in complex and dynamic environments. This paper addresses this challenge by introducing an innovative adaptive task decision-making method, which leverages Deep Q-Network (DQN) algorithms, specifically tailored to meet the unique requirements of heterogeneous UAV swarms. We initially construct a mathematical model that transforms the intricate swarm decision-making issues into a standard Markov Decision Process (MDP), and meticulously design the state space, action space, reward function, and state transition function to cater to the decision-making needs of UAV swarms. Utilizing this model, we further integrate a hybrid experience replay mechanism, coupled with DQN algorithms for offline training, enabling the efficient transformation of training outcomes into a network model that is directly applicable to real-time online decision-making. This approach markedly enhances the decision-making adaptability and dynamic adaptability of UAV swarms when confronted with a variety of tasks. During the offline training phase, we also employ a hybrid experience replay mechanism and target network approximation techniques to bolster the network's stability and training efficiency, ensuring high reliability in online decision-making. Comparative simulations with existing DQN algorithms reveal the distinct superiority of our method in addressing adaptive task decision-making challenges. Moreover, through 1000 comprehensive simulation experiments, we further validate the stability and effectiveness of our method across diverse task environments, providing a solid theoretical foundation and technical support for the practical application and proliferation of UAV swarm technology.

Dissociative and prioritized modeling of behaviorally relevant neural dynamics using recurrent neural networks

Understanding the dynamical transformation of neural activity to behavior requires new capabilities to nonlinearly model, dissociate and prioritize behaviorally relevant neural dynamics and test hypotheses about the origin of nonlinearity. We present dissociative prioritized analysis of dynamics (DPAD), a nonlinear dynamical modeling approach that enables these capabilities with a multisection neural network architecture and training approach. Analyzing cortical spiking and local field potential activity across four movement tasks, we demonstrate five use-cases. DPAD enabled more accurate neural–behavioral prediction. It identified nonlinear dynamical transformations of local field potentials that were more behavior predictive than traditional power features. Further, DPAD achieved behavior-predictive nonlinear neural dimensionality reduction. It enabled hypothesis testing regarding nonlinearities in neural–behavioral transformation, revealing that, in our datasets, nonlinearities could largely be isolated to the mapping from latent cortical dynamics to behavior. Finally, DPAD extended across continuous, intermittently sampled and categorical behaviors. DPAD provides a powerful tool for nonlinear dynamical modeling and investigation of neural–behavioral data.

Layer‐parallel training of residual networks with auxiliary variable networks

AbstractGradient‐based methods for training residual networks (ResNets) typically require a forward pass of input data, followed by back‐propagating the error gradient to update model parameters, which becomes time‐consuming as the network structure goes deeper. To break the algorithmic locking and exploit synchronous module parallelism in both forward and backward modes, auxiliary‐variable methods have emerged but suffer from communication overhead and a lack of data augmentation. By trading off the recomputation and storage of auxiliary variables, a joint learning framework is proposed in this work for training realistic ResNets across multiple compute devices. Specifically, the input data of each processor is generated from its low‐capacity auxiliary network (AuxNet), which permits the use of data augmentation and realizes forward unlocking. Backward passes are then executed in parallel, each with a local loss function derived from the penalty or augmented Lagrangian (AL) method. Finally, the AuxNet is adjusted to reproduce updated auxiliary variables through an end‐to‐end training process. We demonstrate the effectiveness of our method on ResNets and WideResNets across CIFAR‐10, CIFAR‐100, and ImageNet datasets, achieving speedup over the traditional layer‐serial training approach while maintaining comparable testing accuracy.

Vibration based dual-criteria damage detection method using deep neural networks in highway bridges with steel girders

This article introduces a new method for assessing the damage in steel girders of highway bridges using deep neural networks. The method uses modified forms of modal strain energy damage index and modal flexibility damage index to train the neural network and prevent unsafe decisions. The neural network is trained using damage indexes achieved from numerical simulations of the bridge with different damage scenarios. Three types of neural networks, including convolutional neural network (CNN), Long short-term memory (LSTM) neural network, and conventional artificial neural network (ANN) are compared to achieve the best performance. The proposed method overcomes previous detection problems such as false-positive indications, limited damage locations, and scenarios, and insufficient accuracy in cases with multiple damage scenarios. The results show that the proposed method and CNN can accurately identify unspecified damage locations and severity in multi-span highway bridges. The new training method of the CNN deep neural network systems overcomes some shortcomings in ANN, such as being time-consuming, having a low convergence rate, and experiencing data overfitting. Additionally, the CNN training scheme can limit the need for massive amounts of input data and increase the speed and accuracy of network training, especially in multiple damage scenarios.

Are numerical scores important for grant assessment? A cross-sectional study

Background: In the evaluation of research proposals, reviewers are often required to provide their opinions using various forms of quantitative and qualitative criteria. In 2020, the European Commission removed, for the Marie Skłodowska-Curie Actions (MSCA) Innovative Training Networks (ITN) funding scheme, the numerical scores from the individual evaluations but retained them in the consensus report. This study aimed to assess whether there were any differences in reviewer comments’ linguistic characteristics after the numerical scoring was removed, compared to comments from 2019 when numerical scoring was still present. Methods: This was an observational study and the data were collected for the Marie Skłodowska-Curie Actions (MSCA) Innovative Training Networks (ITN) evaluation reports from the calls of 2019 and 2020, for both individual and consensus comments and numerical scores about the quality of the research proposal on three evaluation criteria: Excellence, Impact and Implementation. All comments were analyzed using the Linguistic Inquiry and Word Count (LIWC) program. Results: For both years, the comments for proposal’s strengths were written in a style that reflects objectivity, clout, and positive affect, while in weaknesses cold and objective style dominated, and that pattern remained stable across proposal status and research domains. Linguistic variables explained a very small proportion of the variance of the differences between 2019 and 2020 (McFadden R2=0.03). Conclusions: Removing the numerical scores was not associated with the differences in linguistic characteristics of the reviewer comments. Future studies should adopt a qualitative approach to assess whether there are conceptual changes in the content of the comments.

Towards accurate diagnosis: exploring knowledge distillation and self-attention in multimodal medical image fusion

ABSTRACT Multimodal medical image fusion aims to aggregate significant information based on the characteristics of medical images from different modalities. Existing research in image fusion faces several major limitations, including a scarcity of paired data, noisy and inconsistent modalities, a lack of contextual relationships, and suboptimal feature extraction and fusion techniques. In response to these challenges, this research proposes a novel adaptive fusion approach. Our knowledge distillation (KD) model extracts informative features from multimodal medical images using various key components. A teacher network is employed to emphasise the suitability and complexity of capturing high-level abstract features. The soft labels are utilised to transfer the knowledge between the teacher network as well as the student network. During student network training, we minimise the divergence between these soft labels. To enhance the adaptive fusion of extracted features from different modalities, we apply a self-attention mechanism. Training this self-attention mechanism minimises the loss function, encouraging attention scores to capture relevant contextual relationships between features. Additionally, a cross-modal consistency module aligns the extracted features to ensure spatial consistency and meaningful fusion. Our adaptive fusion strategy effectively combines features to enhance the diagnostic value and quality of fused images. We employ generator and discriminator architectures for synthesising fused images and distinguishing between real and generated fused images. Comprehensive analysis is conducted on the basis of diverse evaluation measures. Experimental results demonstrate improved fusion outcomes with values of 0.92, 41.58, 7.25, 0.958, 0.759, 0.947, 0.90, 7.05, 0.0726, and 76 s for SSIM, PSNR, FF, VIF, UIQI, FMI, EITF, entropy, RMSE, and execution time, respectively.

GSK-LocS: Towards a more effective generalisation in population-based neural network training

Despite the effectiveness of deep neural networks, feed-forward neural networks (FFNNs) continue to play a crucial role in many applications, especially when dealing with limited data availability. The primary challenge in FFNNs is determining the optimal weights during the training process, aiming to minimise the disparity between actual and predicted outputs. Although gradient-based techniques like backpropagation (BP) have traditionally been popular for FFNN training, they come with inherent limitations, such as sensitivity to initial weights and susceptibility to getting trapped in local optima. To overcome these challenges, we introduce a novel approach based on the Gaining-Sharing Knowledge-based(GSK) algorithm. To the best of our knowledge, this paper represents the first exploration of GSK for neural network training. After obtaining the appropriate weights for the FFNN by the GSK, the weights and biases are utilised to initialise a Levenberg–Marquardt backpropagation (LMBP) algorithm, serving as a local search component. In other words, our proposed algorithm, GSK-LocS, leverages the global search capabilities of the GSK algorithm and combines them with the local search capabilities of LMBP for neural network training. This integration mitigates sensitivity to initial values and reduces the risk of being trapped in local optima. Experimental results conducted on classification and approximation problems provide compelling evidence that our proposed algorithm is highly competitive compared to other existing methods.

DropletMask: Leveraging visual data for droplet impact analysis

AbstractMachine learning‐assisted computer vision represents a state‐of‐the‐art technique for extracting meaningful features from visual data autonomously. This approach facilitates the quantitative analysis of images, enabling object detection and tracking. In this study, we utilize advanced computer vision to precisely identify droplet motions and quantify their impact forces with spatiotemporal resolution at the picoliter or millisecond scale. Droplets, captured by a high‐speed camera, are denoised through neuromorphic image processing. These processed images are employed to train convolutional neural networks, allowing the creation of segmented masks and bounding boxes around moving droplets. The trained networks further digitize time‐varying multi‐dimensional droplet features, such as droplet diameters, spreading and sliding motions, and corresponding impact forces. Our innovative method offers accurate measurement of small impact forces with a resolution of approximately 10 pico‐newtons for droplets in the micrometer range across various configurations with the time resolution at hundreds of microseconds.

Improved region proposal network for enhanced few-shot object detection

Despite significant success of deep learning in object detection tasks, the standard training of deep neural networks requires access to a substantial quantity of annotated images across all classes. Data annotation is an arduous and time-consuming endeavor, particularly when dealing with infrequent objects. Few-shot object detection (FSOD) methods have emerged as a solution to the limitations of classic object detection approaches based on deep learning. FSOD methods demonstrate remarkable performance by achieving robust object detection using a significantly smaller amount of training data. A challenge for FSOD is that instances from novel classes that do not belong to the fixed set of training classes appear in the background and the base model may pick them up as potential objects. These objects behave similarly to label noise because they are classified as one of the training dataset classes, leading to FSOD performance degradation. We develop a semi-supervised algorithm to detect and then utilize these unlabeled novel objects as positive samples during the FSOD training stage to improve FSOD performance. Specifically, we develop a hierarchical ternary classification region proposal network (HTRPN) to localize the potential unlabeled novel objects and assign them new objectness labels to distinguish these objects from the base training dataset classes. Our improved hierarchical sampling strategy for the region proposal network (RPN) also boosts the perception ability of the object detection model for large objects. We test our approach and COCO and PASCAL VOC baselines that are commonly used in FSOD literature. Our experimental results indicate that our method is effective and outperforms the existing state-of-the-art (SOTA) FSOD methods. Our implementation is provided as a supplement to support reproducibility of the results https://github.com/zshanggu/HTRPN.11Early partial results of this work is presented in the 2023 ICCV Workshop on Visual Continual Learning (Shangguan and Rostami, 2023).

Analysis of laser beam welding with superimposed 445 and 1070 nm wavelength lasers on copper by in situ synchrotron diagnostics

In laser welding, precision and reproducibility are fundamentally dependent on temporal and spatial processes of energy input. Induced by the dynamics of the melt pool, pressure equilibria in the vapor capillary, and solidification behavior, different weld seam qualities are achieved. To obtain the lowest possible defect frequency, new tailored joining strategies need to be investigated using multibeam and multiwavelength approaches. To improve the quality by influencing the process dynamics, a dual-beam approach is investigated that superimposes a stationary laser beam with a wavelength of 445 nm with a spatially modulated laser beam with a wavelength of 1070 nm. The aim is to utilize ∼10 times higher absorption of a 445 nm diode laser on copper with the high focusability of a 1070 nm fiber laser. In this context, the influence of the relative positions of the two beams to each other on the weld seam quality is investigated, while one of the beams moves either in front, behind, or coaxial to the other beam following the path of a line weld. The main objective is to observe how the laser beams influence each other and how the capillary depth and porosity vary for different parameters. To visualize the process dynamics, the welding experiments on copper are performed at the Deutsches Elektronen-Synchrotron DESY by means of in situ phase contrast videography. Quality-determining weld properties like the distribution of pores or process fluctuations are then extracted automatically from the image sequences by means of a trained neuronal network.

Distilling OCT cervical dataset with evidential uncertainty proxy

Deep learning-based OCT image classification method is of paramount importance for early screening of cervical cancer. For the sake of efficiency and privacy, the emerging data distillation technique becomes a promising way to condense the large-scale original OCT dataset into a much smaller synthetic dataset, without losing much information for network training. However, OCT cervical images often suffer from redundancy, mis-operation and noise, etc. These challenges make it hard to compress as much valuable information as possible into extremely small synthesized dataset. To this end, we design an uncertainty-aware distribution matching based dataset distillation framework (UDM). Precisely, we adopt a pre-trained plug-and-play uncertainty estimation proxy to compute classification uncertainty for each data point in the original and synthetic dataset. The estimated uncertainty allows us to adaptively calculate class-wise feature centers of the original and synthetic data, thereby increasing the importance of typical patterns and reducing the impact of redundancy, mis-operation, and noise, etc. Extensive experiments show that our UDM effectively improves distribution-matching-based dataset distillation under both homogeneous and heterogeneous training scenarios.

Coarse-Grained Molecular Dynamics with Normalizing Flows.

We propose a sampling algorithm relying on a collective variable (CV) of midsize dimension modeled by a normalizing flow and using nonequilibrium dynamics to propose full configurational moves from the proposition of a refreshed value of the CV made by the flow. The algorithm takes the form of a Markov chain with nonlocal updates, allowing jumps through energy barriers across metastable states. The flow is trained throughout the algorithm to reproduce the free energy landscape of the CV. The output of the algorithm is a sample of thermalized configurations and the trained network that can be used to efficiently produce more configurations. We show the functioning of the algorithm first in a test case with a mixture of Gaussians. Then, we successfully tested it on a higher-dimensional system consisting of a polymer in solution with a compact state and an extended stable state separated by a high free energy barrier.

Recognizing Digital Ink Chinese Characters Written by International Students Using a Residual Network with 1-Dimensional Dilated Convolution

Due to the complex nature of Chinese characters, junior international students often encounter writing problems related to strokes, components, and their combinations when writing Chinese characters. Digital ink Chinese characters (DICCs) are obtained by sampling the writing trajectory of Chinese characters with a pen input device. DICCs contain rich information, such as the time and space of strokes and sampling points. Recognizing DICCs is crucial for evaluating and correcting writing errors and enhancing the quality of Chinese character teaching for international students. Here, the paper first employs a one-dimensional dilated convolution to digital ink Chinese character recognition (DICCR) and proposes a novel residual network with one-dimensional dilated convolution (1-D ResNetDC). The 1-D ResNetDC not only utilizes multi-scale convolution kernels, but also employs different dilation rates on a single-scale convolution kernel to obtain information from various ranges. Additionally, residual connections facilitate the training of deep one-dimensional convolutional neural networks. Moreover, the paper proposes a more expressive ten-dimensional feature representation that includes spatial, temporal, and writing direction information for each sampling point, thereby improving classification accuracy. Because the DICC dataset of international students is small and unbalanced, the 1-D ResNetDC is pre-trained on the published available dataset. The experiments demonstrate that our approach is effective and superior. This model features a compact architecture, a reduced number of parameters, and excellent scalability.

Multibeam water column image super-resolution by combining morphological and intensity features

ABSTRACT Multibeam water column images (WCI) play a crucial role in the detection and recognition of underwater targets. However, WCI is not a complete image of underwater space due to the gaps between beams inherent in multibeam systems. In this study, we design the WCI super-resolution network (WCISRN) based on Swin-transformer to adapt to the uneven and sparse distribution of water column data. To enable WCISRN to learn morphological features effectively, we produced simulated datasets for pre-training. Subsequently, we conduct self-supervised training on the measured WCI dataset to fine-tune the network parameters, aiming to better adapt WCISRN to the backscattering intensity features of WCI. Finally, by performing network cross-fusion, the pre-trained network and self-supervised training network are combined to form a new network that achieves a balance between image morphology and intensity features. The proposed method is evaluated through visual analysis and quantitative comparison using simulated and real-world datasets. The experimental results demonstrate that our method exhibits stronger detail restoration ability and clearer image morphology. The average peak signal-to-noise ratio (PSNR) on the test set improved from 23.75–26.14, indicating a significant enhancement in the imaging quality of WCI.

Form-finding of tensegrity structures based on graph neural networks

Tensegrity structures, characterized by enhanced stiffness, slender struts, and superior buckling resistance, have found wide-ranging applications in fields such as engineering, architecture, art, biology, and robotics, attracting extensive attention from researchers. The form-finding process, a critical step in the design of tensegrity structures, aims to discover the self-equilibrated configuration that satisfies specific design requirements. Traditional form-finding methods based on force density often require repeated steps of eigenvalue decomposition and singular value decomposition, making the process complex. In contrast, this paper introduces a new intelligent form-finding algorithm that uses the force density method and combines the Coati optimization algorithm with Graph Neural Networks. This algorithm avoids the complex steps of eigenvalue and singular value decomposition and integrates the physical knowledge of the structure, making the form-finding process faster and more accurate. In this algorithm, various force densities are initially randomized and input into a trained Graph Neural Networks to predict a fitness function’s value. Through optimizing the constrained fitness function, the algorithm determines the appropriate structural force density and coordinates, thereby completing the form-finding process of the structure. The paper presents seven typical tensegrity structure examples and compares various form-finding methods. The results of numerical examples show that the method proposed in this paper can find solutions that align with the super-stable line more quickly and accurately, demonstrating its potential value in practical applications.

Dynamic data reconciliation for enhancing the prediction performance of long short-term memory network

Abstract In modern process industries, long short-term memory (LSTM) network is widely used for data-driven modeling. Constrained by measuring instruments and environments, the measured datasets are generally with Gaussian/non-Gaussian distributed measurement noise. The noisy datasets will impact the modeling accuracy of the LSTM network and decrease the prediction performance of it. Aiming at addressing prediction performance impairment of the LSTM network under noisy datasets with Gaussian/non-Gaussian distribution, this study introduces dynamic data reconciliation (DDR) both into LSTM network training and into LSTM network test. Results show that DDR improves not only the data quality based on noisy datasets and the training outputs via the Bayesian formula in the model training step, but also the prediction performance based on offline measured information and the test outputs. The implementation scheme of DDR for Gaussian and non-Gaussian distributed noise is purposely designed. The effectiveness of DDR on the LSTM model is verified in a numerical example and a case involving a set of shared wind power datasets.

Super-Resolving and Denoising 4D flow MRI of Neurofluids Using Physics-Guided Neural Networks.

To obtain high-resolution velocity fields of cerebrospinal fluid (CSF) and cerebral blood flow by applying a physics-guided neural network (div-mDCSRN-Flow) to 4D flow MRI. The div-mDCSRN-Flow network was developed to improve spatial resolution and denoise 4D flow MRI. The network was trained with patches of paired high-resolution and low-resolution synthetic 4D flow MRI data derived from computational fluid dynamic simulations of CSF flow within the cerebral ventricles of five healthy cases and five Alzheimer's disease cases. The loss function combined mean squared error with a binary cross-entropy term for segmentation and a divergence-based regularization term for the conservation of mass. Performance was assessed using synthetic 4D flow MRI in one healthy and one Alzheimer' disease cases, an in vitro study of healthy cerebral ventricles, and in vivo 4D flow imaging of CSF as well as flow in arterial and venous blood vessels. Comparison was performed to trilinear interpolation, divergence-free radial basis functions, divergence-free wavelets, 4DFlowNet, and our network without divergence constraints. The proposed network div-mDCSRN-Flow outperformed other methods in reconstructing high-resolution velocity fields from synthetic 4D flow MRI in healthy and AD cases. The div-mDCSRN-Flow network reduced error by 22.5% relative to linear interpolation for in vitro core voxels and by 49.5% in edge voxels. The results demonstrate generalizability of our 4D flow MRI super-resolution and denoising approach due to network training using flow patches and physics-based constraints. The mDCSRN-Flow network can facilitate MRI studies involving CSF flow measurements in cerebral ventricles and association of MRI-based flow metrics with cerebrovascular health.

Neural force functional for non-equilibrium many-body colloidal systems

Abstract We combine power functional theory and machine learning to study non-equilibrium overdamped many-body systems of colloidal particles at the level of one-body fields. We first sample in steady state the one-body fields relevant for the dynamics from computer simulations of Brownian particles under the influence of randomly generated external fields. A neural network is then trained with this data to represent locally in space the formally exact functional mapping from the one-body density and velocity profiles to the one-body internal force field. The trained network is used to analyse the non-equilibrium superadiabatic force field and the transport coefficients such as shear and bulk viscosities. Due to the local learning approach, the network can be applied to systems much larger than the original simulation box in which the one-body fields are sampled. Complemented with the exact non-equilibrium one-body force balance equation and a continuity equation, the network yields viable predictions of the dynamics in time-dependent situations. Even though training is based on steady states only, the predicted dynamics is in good agreement with simulation results. A neural dynamical density functional theory can be straightforwardly implemented as a limiting case in which the internal force field is that of an equilibrium system. The framework is general and directly applicable to other many-body systems of interacting particles following Brownian dynamics.

Performance Analysis of Cooperative Spectrum Sensing using Empirical Mode Decomposition and Artificial Neural Network in Wireless Regional Area Network

Background: Radio spectrum is natural and the most precious means in wireless communication systems. Optimal spectrum utilization is a key concern for today's cutting-edge wireless communication networks. The impending problem of the lack of available spectrum has prompted the development of a new idea called “Cognitive Radio” (CR). Cooperative spectrum sensing (CSS) is utilized to improve the detection performance of the system. Several fusion algorithms of decision-making are proposed for sensing the licensed user, but they do not work well under low signal-to-noise ratio (SNR). Objectives: To address the issue of poor detection performance under low SNR, Empirical mode decomposition (EMD) and artificial neural network (ANN) based CSS under Rayleigh multipath fading channel in IEEE 802.22 wireless regional area network (WRAN) is proposed in this paper. Method: In this work, we propose the use of ANN as a fusion center. First, the received signal's energy is calculated using EMD. The computed energy, SNR, and false alarm probability are combined to form a data set of 2048 samples. They are utilized to train Levenberg- Marquardt back propagation training algorithm-based feed-forward neural network (FFNN). Using this trained network, CSS in WRAN is simulated under Rayleigh multipath fading. Results: Simulation results show that the proposed CSS method based on EMD-ANN outperforms the standard fast Fourier transform (FFT) and EMD detection-based cooperative spectrum sensing with a hard "OR" fusion at low SNR. With Pf =0.01, 100% detection accuracy with proposed techniques is obtained at SNR= -22dB. Conclusion: The findings show that the suggested approach outperforms EMD and FFT based energy detection scheme-based traditional CSS in low SNR environments.

An Effective Res-Progressive Growing Generative Adversarial Network-Based Cross-Platform Super-Resolution Reconstruction Method for Drone and Satellite Images

In recent years, accurate field monitoring has been a research hotspot in the domains of aerial remote sensing and satellite remote sensing. In view of this, this study proposes an innovative cross-platform super-resolution reconstruction method for remote sensing images for the first time, aiming to make medium-resolution satellites capable of field-level detection through a super-resolution reconstruction technique. The progressive growing generative adversarial network (PGGAN) model, which has excellent high-resolution generation and style transfer capabilities, is combined with a deep residual network, forming the Res-PGGAN model for cross-platform super-resolution reconstruction. The Res-PGGAN architecture is similar to that of the PGGAN, but includes a deep residual module. The proposed Res-PGGAN model has two main benefits. First, the residual module facilitates the training of deep networks, as well as the extraction of deep features. Second, the PGGAN structure performs well in cross-platform sensor style transfer, allowing for cross-platform high-magnification super-resolution tasks to be performed well. A large pre-training dataset and real data are used to train the Res-PGGAN to improve the resolution of Sentinel-2’s 10 m resolution satellite images to 0.625 m. Three evaluation metrics, including the structural similarity index metric (SSIM), the peak signal-to-noise ratio (PSNR), and the universal quality index (UQI), are used to evaluate the high-magnification images obtained by the proposed method. The images generated by the proposed method are also compared with those obtained by the traditional bicubic method and two deep learning super-resolution reconstruction methods: the enhanced super-resolution generative adversarial network (ESRGAN) and the PGGAN. The results indicate that the proposed method outperforms all the comparison methods and demonstrates an acceptable performance regarding all three metrics (SSIM/PSNR/UQI: 0.9726/44.7971/0.0417), proving the feasibility of cross-platform super-resolution image recovery.