Convolutional Autoencoder Model Research Articles

Increasing Offline Handwritten Chinese Character Recognition Using Separated Pre-Training Models: A Computer Vision Approach

Offline handwritten Chinese character recognition involves the application of computer vision techniques to recognize individual handwritten Chinese characters. This technology has significantly advanced the research in online handwriting recognition. Despite its widespread application across various fields, offline recognition faces numerous challenges. These challenges include the diversity of glyphs resulting from different writers’ styles and habits, the vast number of Chinese character labels, and the presence of morphological similarities among characters. To address these challenges, an optimization method based on a separated pre-training model was proposed. The method aims to enhance the accuracy and robustness of recognizing similar character images by exploring potential correlations among them. In experiments, the HWDB and Chinese Calligraphy Styles by Calligraphers datasets were employed, utilizing precision, recall, and the Macro-F1 value as evaluation metrics. We employ a convolutional self-encoder model characterized by high recognition accuracy and robust performance. The experimental results demonstrated that the separated pre-training models improved the performance of the convolutional auto-encoder model, particularly in handling error-prone characters, resulting in an approximate 6% increase in precision.

Automated fabric inspection system development aided with convolutional autoencoder-based defect detection

Industrial automatic fabric inspection system, a critical technology in the industry, enhances both total production quantity and quality compared to conventional inspection techniques. This study aims to create a reliable and effective real-time automated visual inspection system for fabrics, focusing on defect detection. The goals of the study can be stated as; installing a system with advanced technology for capturing and processing images swiftly, the development and deployment of a system capable of autonomously learning and scanning fabrics in use, and the creation of a smart framework for accurate fabric defect detection and classification. We focus on the development of unsupervised fabric defect detection using a convolutional autoencoder model, and defect classification using a convolutional neural network model, which takes input as the feature vector generated by the convolutional autoencoder. The experimental outcomes have displayed significant success rates in both detecting defects and classifying them, confirming the effectiveness of the framework in real-time visual inspection systems.

Deep learning model for fiber optic distributed acoustic sensing seismic data compression based on autoencoder and recurrent neural networks

Because of the extremely high sampling rate of fiber optic distributed acoustic sensing (DAS) equipment, the amount of DAS seismic data collected is enormous, which poses great challenges to the transmission and storage of DAS seismic data. Therefore, it is essential to study the compression methods of DAS seismic data. Existing data compression methods such as wavelet transform, cosine transform, and convolutional autoencoder (CAE) still have room for improvement in the compression performance and compression ratio (CR). Thus, we have proposed what we believe to be is a novel deep learning model called the recurrent autoencoder (RAE) for high-performance compression of DAS seismic data. Under different CRs, we have designed performance evaluation experiments for RAE models based on different RNN modules with different loss functions. When the CR of the RAE model is 8, the signal-to-noise ratio (SNR) of the reconstructed DAS seismic data reaches 40.60 dB, which is better than that of the CAE’s 11.58 dB. The ultimate CR was increased to 512 without reducing the compression quality, which is 4.12 times higher than the CAE model. The LSTM with a weighted loss function improves the SNR to 43.66 dB at a CR of 8, which is 3.06 dB higher than the LSTM with additive loss function. The results show that the RAE model proposed with a weighted loss function in this paper has excellent DAS seismic data compression performance and provides a high CR, which can be widely applied in large-scale DAS seismic data compression.

Convolutional Autoencoder for Reconstruction of Historical Document Images: Ancient Manuscript Babad Lombok

The Babad Lombok is an ancient literary or manuscripts document that generally contains stories about the origins of the people of Lombok. This document is written on a lontar leaf, which in the past was used to write manuscripts, letters, and documents. At present, the Babad Lombok document can be seen in the form of photos or scans, so it can be viewed without having to go to a museum or cultural heritage site where the document is usually exhibited. However, because this document is an ancient artifact that has been around for hundreds of years, it has naturally experienced fading in the original document or its scanned versions. This makes the text inside less clear. This paper proposes to automatically reconstruct/repair the Babad Lombok document using a neural network. The type of neural network used is an Autoencoder or Convolutional Autoencoder (CAE). The CAE model is built sequentially and trained using original images of Babad Lombok as its training data and manually corrected images of Babad Lombok as the target or ground truth data. In the process, the two types of data are iteratively cropped to a size of 64x64 along the original size of the Babad Lombok image. This process results in input and target data for the CAE training process in this research, each consisting of 48,288 images. Testing the trained autoencoder model shows that the Babad images have been successfully repaired, making the text quality clearer before reconstruction. Ultimately, the proposed CAE has achieved training and validation accuracies of 89.09% and 94.57%, with corresponding loss values of 0.0418 and 0.0226.

Unsupervised deep representation learning enables phenotype discovery for genetic association studies of brain imaging

Understanding the genetic architecture of brain structure is challenging, partly due to difficulties in designing robust, non-biased descriptors of brain morphology. Until recently, brain measures for genome-wide association studies (GWAS) consisted of traditionally expert-defined or software-derived image-derived phenotypes (IDPs) that are often based on theoretical preconceptions or computed from limited amounts of data. Here, we present an approach to derive brain imaging phenotypes using unsupervised deep representation learning. We train a 3-D convolutional autoencoder model with reconstruction loss on 6130 UK Biobank (UKBB) participants’ T1 or T2-FLAIR (T2) brain MRIs to create a 128-dimensional representation known as Unsupervised Deep learning derived Imaging Phenotypes (UDIPs). GWAS of these UDIPs in held-out UKBB subjects (n = 22,880 discovery and n = 12,359/11,265 replication cohorts for T1/T2) identified 9457 significant SNPs organized into 97 independent genetic loci of which 60 loci were replicated. Twenty-six loci were not reported in earlier T1 and T2 IDP-based UK Biobank GWAS. We developed a perturbation-based decoder interpretation approach to show that these loci are associated with UDIPs mapped to multiple relevant brain regions. Our results established unsupervised deep learning can derive robust, unbiased, heritable, and interpretable brain imaging phenotypes.

APPLICATION OF OBJECT DETECTION TECHNOLOGY IN AORTOGRAPHY KEYPOINT TRACKING

HighlightsThe practical significance of the work lies in the fact that the presented aortography data visualization system is an effective tool for visually assisting surgeons during transcatheter aortic valve implantation interventions, supporting real-time operation mode. The proposed data preprocessing algorithm, which improves image quality with minimal performance costs, complements the system, allowing specialists to achieve the best result. AbstractAim. The aim of this study is to develop a visual assistance system for transcatheter aortic valve implantation procedures.Methods. To address the stated objective, our own dataset consisting of 35 videos of the intervention was used. The visualization system is based on the approach of detecting key points in aortography, utilizing “Object detection” technology with the application of YOLO family artificial neural networks. To achieve the best result, we proposed a method to enhance the quality of input data using convolutional neural networks, specifically the «Autoencoder» technology.Results. The results of the study revealed that the convolutional autoencoder model is capable of restoring the informativeness of noisy input images from 40 to 75%, thereby increasing the accuracy of object detection in images. The presented real-time tracking system for facilitating TAVI procedures achieves a final accuracy of 51.9% according to the Mean Average Precision (MAP) quality metric. The visual assistance system can recognize and track key points indicating the location of the aortic root, delivery system, and heart valve prosthesis during surgery. The practical significance of the work lies in the fact that the presented aortography data visualization system is an effective tool for visually assisting surgeons during interventions, supporting real-time operation mode.Conclusion. The proposed data preprocessing algorithm, which improves image quality with minimal performance costs, complements the visualization system, allowing specialists to achieve the best results.

Open Set Bearing Fault Diagnosis with Domain Adaptive Adversarial Network under Varying Conditions

Bearing fault diagnosis is a pivotal aspect of monitoring rotating machinery. Recently, numerous deep learning models have been developed for intelligent bearing fault diagnosis. However, these models have typically been established based on two key assumptions: (1) that identical fault categories exist in both the training and testing datasets, and (2) the datasets used for testing and training are assumed to follow the same distribution. Nevertheless, these assumptions prove impractical and fail to accurately depict real-world scenarios, particularly those involving open-world assumption fault diagnosis in multi-condition scenarios. For that purpose, an open set domain adaptive adversarial network framework is proposed. Specifically, in order to improve the learning of distribution characteristics in different fields, comprehensive training is implemented using a deep convolutional autoencoder model. Additionally, to mitigate the negative transfer resulting from unknown fault samples in the target domain, the similarity of each target domain sample and the shared classes in the source domain are estimated using known class classifiers and extended classifiers. Similarity weight values are assigned to each target domain sample, and an unknown boundary is established in a weighted manner. This approach is employed to establish the alignment between the classes shared between the two domains, enabling the classification of known fault classes, while allowing the recognition of unknown fault classes in the target domain. The efficacy of our suggested approach is empirically validated using different datasets.

Interference Suppression for an FM-Radio-Based Passive Radar via Deep Convolutional Autoencoder

Recently, the passive bistatic radar (PBR) which exploits frequency modulation (FM) radio transmitters as illuminators, has witnessed widespread usage owing to its various advantages. However, the characteristics of FM-radio-based PBR result in interference components in the range–Doppler (RD) map, which may increase false alarms. Therefore, this study proposes a method for suppressing interference components using a deep learning approach. The two main contributions of this study are as follows. First, a convolutional autoencoder model capable of effectively suppressing interference in the RD map of PBR was proposed. Second, a synthetic RD map dataset generation method that can enable the autoencoder to operate robustly in PBR in a real environment was presented. Further, a performance comparison between the proposed method and existing methods using simulated data proved that the deep learning-based method exhibited superior target detection performance. Furthermore, using the data recorded by the PBR in a real environment, the proposed autoencoder model was shown to effectively suppress interference components in a real interference environment.

ECG based one-dimensional residual deep convolutional auto-encoder model for heart disease classification

ECG based one-dimensional residual deep convolutional auto-encoder model for heart disease classification

PDSE-Lite: lightweight framework for plant disease severity estimation based on Convolutional Autoencoder and Few-Shot Learning.

Plant disease diagnosis with estimation of disease severity at early stages still remains a significant research challenge in agriculture. It is helpful in diagnosing plant diseases at the earliest so that timely action can be taken for curing the disease. Existing studies often rely on labor-intensive manually annotated large datasets for disease severity estimation. In order to conquer this problem, a lightweight framework named "PDSE-Lite" based on Convolutional Autoencoder (CAE) and Few-Shot Learning (FSL) is proposed in this manuscript for plant disease severity estimation with few training instances. The PDSE-Lite framework is designed and developed in two stages. In first stage, a lightweight CAE model is built and trained to reconstruct leaf images from original leaf images with minimal reconstruction loss. In subsequent stage, pretrained layers of the CAE model built in the first stage are utilized to develop the image classification and segmentation models, which are then trained using FSL. By leveraging FSL, the proposed framework requires only a few annotated instances for training, which significantly reduces the human efforts required for data annotation. Disease severity is then calculated by determining the percentage of diseased leaf pixels obtained through segmentation out of the total leaf pixels. The PDSE-Lite framework's performance is evaluated on Apple-Tree-Leaf-Disease-Segmentation (ATLDS) dataset. However, the proposed framework can identify any plant disease and quantify the severity of identified diseases. Experimental results reveal that the PDSE-Lite framework can accurately detect healthy and four types of apple tree diseases as well as precisely segment the diseased area from leaf images by using only two training samples from each class of the ATLDS dataset. Furthermore, the PDSE-Lite framework's performance is compared with existing state-of-the-art techniques, and it is found that this framework outperformed these approaches. The proposed framework's applicability is further verified by statistical hypothesis testing using Student t-test. The results obtained from this test confirm that the proposed framework can precisely estimate the plant disease severity with a confidence interval of 99%. Hence, by reducing the reliance on large-scale manual data annotation, the proposed framework offers a promising solution for early-stage plant disease diagnosis and severity estimation.

Early neurological deterioration detection with a transformer convolutional auto-encoder model

This paper proposes an improved transformer convolutional auto-encoder model combined with the exponentially weighted moving average (EWMA) control chart to detect early neurological deterioration (END) of ischemic stroke patients after endovascular therapy in advance. In the proposed method, the transformer convolutional auto-encoder is used to extract crucial features of multivariate clinical monitoring time series data and obtain the reconfiguration loss while EWMA control chart is utilized to monitor the derived reconfiguration loss and identify anomalies. To verify the feasibility and effectiveness of the proposed END detection approach, multivariate clinical monitoring time series data of ischemic stroke patients in the neurocritical care unit from Beijing Tiantan hospital are collected. Meanwhile, the proposed approach is benchmarked with seven state-of-the-art models. The computation results show that the proposed approach achieves the best performance with the lowest false alarm rate and the highest detection rate. Therefore, the proposed END detection model is practical to guide doctors in conducting clinical interventions in advance to prevent deterioration in patients with ischemic stroke.

Brain imaging endophenotypes by unsupervised deep learning

AbstractBackgroundUnderstanding the genetic architecture of brain structure and Alzheimer’s disease is challenging, partly due to difficulties in designing robust, non‐biased descriptors of the brain.MethodWe present approaches to derive robust brain imaging phenotypes using unsupervised deep representation learning. Training a 3‐D convolutional autoencoder model with reconstruction loss, we derived a vector representation (termed endophenotype) that captures rich morphological information of the brain. Further, we developed a perturbation‐based decoder interpretation approach that can highlight brain regions that are most relevant to individual endophenotypes.ResultUsing UK Biobank data, we trained the model using over 6,000 UK Biobank (UKBB) participants’ T1 or T2‐FLAIR (T2) brain MRIs. The model is used to derive 128 endophenotypes. The endophenotypes have a mean heritability of 0.3 and the GWAS of which have identified 43 independent loci in the held‐out UKBiobank dataset (discovery n = 22,962/replication n = 12,848), among which 13 loci have not been previously reported by the UK Biobank Big40 study.ConclusionUsing UK Biobank data, we show that, compared to traditional brain image‐derived phenotypes (IDPs), our endophenotypes are more heritable and have higher power for genetic discovery.

Dispersion compensation of Lamb waves based on a convolutional auto-encoder

Lamb wave-based detection has become a promising technique for structural health monitoring in plate-like structures. However, the dispersion effect of Lamb waves makes the wave packets elongated, which degrades the resolution for damage identification. Hence, it is necessary to develop effective methods for the dispersion removal of Lamb waves. Recently, deep learning has drawn much attention for Lamb wave detection due to its powerful feature extraction capability. Most current deep learning models are developed directly for damage classification or location estimation tasks on the Lamb wave signals, rather than signal processing for further manipulation. Therefore, in this paper, a novel approach based on a convolutional auto-encoder is proposed for the dispersion compensation of Lamb waves. The convolutional auto-encoder model is utilized to construct the mapping relationship between the dispersive signal and the time of flight of the wave packet. The dispersion compensated signal is reconstructed by combining the estimated time of flight from the network with the waveform of the excitation signal. Numerical and experimental validations on aluminum and composite plates are implemented to verify the effectiveness of the proposed method. Compared with conventional methods for dispersion compensation, the results demonstrate that our proposed method not only separates the overlapped wave packets but also enables dispersion compensation of multimodal and multi-packet Lamb wave signals. In addition, this method is still applicable in the case of inaccurate dispersion data due to the generalizability of the data-driven model.

A noise-reduction method for UUV communication using a stack-type convolutional autoencoder model of image features

Underwater acoustic (UWA) communication is the primary interaction between UUV and underwater array. However, noise can seriously affect the quality of communication. The conventional filtering algorithms and multiple kernel learning (MKL) models are unsuitable for abrupt-varying noise environments. To address the problem, this study proposes a deep learning (DL) noise-reduction model based on image features. Firstly, in the preprocessing stage, we transform the abrupt-varying noise into a two-dimensional representation of the image characteristics and construct a time–frequency preprocessing structure based on higher-order DEMON spectral representation. Secondly, a feature extraction method based on EMD feature enhancement is built to improve computing efficiency. Finally, we propose a noise-reduction model based on a stack-type convolutional autoencoder aiming to suppress underwater noise and enhance the system’s communication quality. The simulations and experiments on a lake show that the proposed method is more robust over the changing signal-to-noise ratio (SNR) and has a lower bit error rate (BER) than conventional methods.

A novel unsupervised anomaly detection method for rotating machinery based on memory augmented temporal convolutional autoencoder

During the operation of rotating machinery, the occurrence of unknown fault types makes it impossible for the artificial intelligence-based fault diagnosis model to distinguish. Furthermore, due to the excessive generalization capability of the autoencoder, the unsupervised anomaly detection method based on the autoencoder is difficult to effectively distinguish normal and abnormal samples. To address the above problem, this paper proposed an unsupervised anomaly detection method based on memory augmented temporal convolutional autoencoder (MATCAE). Firstly, a novel temporal convolutional autoencoder model is constructed based on dilated causal convolution, skip connection and autoencoder to facilitate the model to learn the temporal features of the input data, thereby enhancing the model’s ability to capture the complex structure of the data. Then, a memory augmented module is designed using a memory matrix and an attention mechanism to expand the distribution interval between the reconstructed samples of normal and abnormal samples and reduce the sample capacity in the overlapping area. Finally, an anomaly detection module based on Euclidean distance, cosine distance and absolute mean square error is designed to improve the reliability of the metric between the input and reconstructed samples. To verify the effectiveness of the proposed method, experimental validation is carried out on a gearbox anomaly detection dataset. The experimental results show that the proposed method has higher anomaly detection accuracy and better noise robustness than other advanced anomaly detection methods, where the average performance metric is improved by 26.86% at the highest and 2.80% at the lowest.

Assessment of the effectiveness of a convolutional autoencoder for digital image-based automated core logging

Widely applicable convolutional neural network (CNN)-based lithology classification models are limited to interpret soundness of a trained model and require high computational cost. This study proposes a convolutional autoencoder (CAE)-based classification model to improve the efficiency of rock-core-logging analysis based on digital rock-core images (CAE model I) and suggest a more understandable classification model by extracting the lithological feature information from the network structure of the model. Moreover, input digital-image data transformation using the Canny edge-detection algorithm was implemented (CAE model II). The applicability of the developed models is validated by comparing them with a CNN model and RGBE model. While implementing the proposed model, digital images of 85-m-long cores were used, which were acquired from the Satyr 5 well in the Northern Carnarvon Basin, Western Australia. It was found that CAE models I and II showed classification accuracy comparable to the CNN model (CAE model I: 96.4%, CAE model II: 94.9%, CNN model: 97.5%, and RGBE model: 94.4%). The extracted features of the trained CAE models are effective in interpreting the trained network structure. In addition, in terms of computational efficiency and training time, the CAE-based models are significantly better than the CNN model (CAE model I: 7 min 21 s, CAE model II: 7 min 34 s, CNN model: 12 min 6 s, and RGBE model: 24 min 35 s). Overall, the results demonstrate the feasibility of the proposed CAE-based models for automated rock-core logging.

Content-based image retrieval for the diagnosis of myocardial perfusion imaging using a deep convolutional autoencoder

BackgroundSingle-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI) plays a crucial role in the optimal treatment strategy for patients with coronary heart disease. We tested the feasibility of feature extraction from MPI using a deep convolutional autoencoder (CAE) model. MethodsEight hundred and forty-three pairs of stress and rest myocardial perfusion images were collected from consecutive patients who underwent cardiac scintigraphy in our hospital between December 2019 and February 2022. We trained a CAE model to reproduce the input paired image data, so as the encoder to output a 256-dimensional feature vector. The extracted feature vectors were further dimensionally reduced via principal component analysis (PCA) for data visualization. Content-based image retrieval (CBIR) was performed based on the cosine similarity of the feature vectors between the query and reference images. The agreement of the radiologist’s finding between the query and retrieved MPI was evaluated using binary accuracy, precision, recall, and F1-score. ResultsA three-dimensional scatter plot with PCA revealed that feature vectors retained clinical information such as percent summed difference score, presence of ischemia, and the location of scar reported by radiologists. When CBIR was used as a similarity-based diagnostic tool, the binary accuracy was 81.0%. ConclusionThe results indicated the utility of unsupervised feature learning for CBIR in MPI. Graphical abstract▪

RcCAE: a convolutional autoencoder method for detecting intra-tumor heterogeneity and single-cell copy number alterations.

Intra-tumor heterogeneity (ITH) is one of the major confounding factors that result in cancer relapse, and deciphering ITH is essential for personalized therapy. Single-cell DNA sequencing (scDNA-seq) now enables profiling of single-cell copy number alterations (CNAs) and thus aids in high-resolution inference of ITH. Here, we introduce an integrated framework called rcCAE to accurately infer cell subpopulations and single-cell CNAs from scDNA-seq data. A convolutional autoencoder (CAE) is employed in rcCAE to learn latent representation of the cells as well as distill copy number information from noisy read counts data. This unsupervised representation learning via the CAE model makes it convenient to accurately cluster cells over the low-dimensional latent space, and detect single-cell CNAs from enhanced read counts data. Extensive performance evaluations on simulated datasets show that rcCAE outperforms the existing CNA calling methods, and is highly effective in inferring clonal architecture. Furthermore, evaluations of rcCAE on two real datasets demonstrate that it is able to provide a more refined clonal structure, of which some details are lost in clonal inference based on integer copy numbers.

Detection of protein, starch, oil, and moisture content of corn kernels using one-dimensional convolutional autoencoder and near-infrared spectroscopy.

Analysis of the nutritional values and chemical composition of grain products plays an essential role in determining the quality of the products. Near-infrared spectroscopy has attracted the attention of researchers in recent years due to its advantages in the analysis process. However, preprocessing and regression models in near-infrared spectroscopy are usually determined by trial and error. Combining newly popular deep learning algorithms with near-infrared spectroscopy has brought a new perspective to this area. This article presents a new method that combines a one-dimensional convolutional autoencoder with near-infrared spectroscopy to analyze the protein, moisture, oil, and starch content of corn kernels. First, a one-dimensional convolutional autoencoder model was created for three different spectra in the corn dataset. Thirty-two latent variables were obtained for each spectrum, which is a low-dimensional spectrum representation. Multiple linear regression models were built for each target using the latent variables of obtained autoencoder models. R2, RMSE, and RMSPE were used to show the performance of the proposed model. The created one-dimensional convolutional autoencoder model achieved a high reconstruction rate with a mean RMSPE value of 1.90% and 2.27% for calibration and prediction sets, respectively. This way, a spectrum with 700 features was converted to only 32 features. The created MLR models which use these features as input were compared to partial least squares regression and principal component regression combined with various preprocessing methods. Experimental results indicate that the proposed method has superior performance, especially in MP5 and MP6 datasets.

A convolutional autoencoder model with weighted multi-scale attention modules for 3D skeleton-based action recognition

The 3D skeleton sequences of action can be recognized based on series of meaningful movements including changes in the direction and geometry features of the body pose. In this paper, we introduce the 3DPo-CDP descriptor, which incorporates the change direction patterns of body joints and pose features in a unified deep structure to learn more adequate features for the action recognition problem. To this end, two types of features are extracted. First, Change Direction Patterns (CDPs) are extracted by following the important points of motion trajectories where a significant change of direction has occurred using two filtering phases. The CDPs capture the global features which are invariant to noise and insignificant temporal dynamics of joints. Second, Pose Features are employed to learn the intrinsic connectivity relationships of adjacent limbs and the variance distances of body joints from representative joints to concentrate on key-frames and informative joints. The complementary features of CDPs and 3D pose, which are transformed into images, are combined in a unified representation and fed into a new convolutional autoencoder. Unlike conventional convolutional autoencoders that focus on frames, high-level discriminative features of spatiotemporal relationships of whole body joints are extracted by introducing weighted multi-scale channel and spatial attention modules. In this paper, we show that adjacent and non-adjacent neighbors can be effectively used to compute different weights for extracting cross-interaction channels and multi-scale spatial relationships of the current pixel. The extracted features are combined with the wavelet representation of statistical body information and then classified with a multi-class SVM classifier. The experimental results demonstrate the effectiveness of the augmented 3DPo-CDP descriptor using an attentional convolution autoencoder structure on five challenging 3D action recognition datasets.