Deep Learning Processing Research Articles

Leveraging Fine-Tuned Language Models in Bioinformatics: A Research Perspective

Bioinformatics, an interdisciplinary field combining biology, computer science, and statistics, has advanced with deep learning and natural language processing techniques. This perspective explores the applications of fine-tuned language models in bioinformatics, highlighting their potential in various domains while discussing challenges and limitations. Fine-tuned language models benefit biomedical literature analysis, extracting information from scientific papers to synthesize knowledge and generate synthetic sequences for DNA, RNA, and protein research. In drug discovery, these models can identify novel drug targets, accelerate virtual screening, and aid drug repurposing by finding new therapeutic indications for existing drugs. For clinical decision support, fine-tuned language models can analyse patient data, medical literature, and guidelines to provide personalized recommendations and alerts to healthcare professionals. They can also aid accurate protein structure prediction for drug design and target identification. In pharmacovigilance, these models can analyse unstructured data sources to detect adverse events from social media, patient forums, and health records, enabling early intervention and improving patient safety. However, challenges like data availability, domain-specific knowledge, bias, interpretability, resource efficiency, ethics, and validation must be addressed for reliable application. Addressing these challenges will unlock the full potential of fine-tuned language models in bioinformatics, driving advancements and benefiting human health. Collaboration between computational and experimental biologists, ethicists, and regulatory bodies is crucial to establish ethical guidelines and best practices for their use.

Online Failure Detection using Deep Learning in FPGA PCB Interface

This research paper is aimed to present a real-time failure detection technique while working with Field Programmable Gate Arrays (FPGA) and interfaced Printed Circuit Boards (PCBs). In this research, we explored the feasibility of currently available innovative Deep Learning (DL) algorithms to detect the defects in variety of PCBs. In our proposed technique, we trained the YOLOv5 (You Only Look Once) algorithm with a few hundreds of defective PCBs’ images, which were obtained from Kaggle, an online community of data scientists and machine learning practitioners. The advantage of using YOLOv5 is that the detection is carried out in real-time. In the next phase, after training, the algorithm undergoes validation and testing, where we tested with different images. The obtained results are promising, as the Deep Learning process successfully detects the defects on the PCBs.

Information Technology Usage in Skin Disease Detection

Millions of individuals of all ages are affected by skin diseases, a widespread problem worldwide. Early diagnosis and detection are essential for these diseases to be effectively treated and improve patient outcomes. Automated skin disease detection systems are a viable way to increase diagnostic accuracy and lighten the workload of dermatologists, by developments in machine learning and computer vision. These systems examine skin lesions and categorize them into several disease groups using various techniques, including feature extraction, deep learning, and image processing. Such systems are still being developed to enhance their precision and usefulness. This paper provides an overview of the different information technologies in skin disease detection, including their effectiveness, the challenges and limitations of existing systems, and future research directions in this field.

Generation of Conventional 18F-FDG PET Images from 18F-Florbetaben PET Images Using Generative Adversarial Network: A Preliminary Study Using ADNI Dataset.

Background and Objectives: 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET) (PETFDG) image can visualize neuronal injury of the brain in Alzheimer's disease. Early-phase amyloid PET image is reported to be similar to PETFDG image. This study aimed to generate PETFDG images from 18F-florbetaben PET (PETFBB) images using a generative adversarial network (GAN) and compare the generated PETFDG (PETGE-FDG) with real PETFDG (PETRE-FDG) images using the structural similarity index measure (SSIM) and the peak signal-to-noise ratio (PSNR). Materials and Methods: Using the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, 110 participants with both PETFDG and PETFBB images at baseline were included. The paired PETFDG and PETFBB images included six and four subset images, respectively. Each subset image had a 5 min acquisition time. These subsets were randomly sampled and divided into 249 paired PETFDG and PETFBB subset images for the training datasets and 95 paired subset images for the validation datasets during the deep-learning process. The deep learning model used in this study is composed of a GAN with a U-Net. The differences in the SSIM and PSNR values between the PETGE-FDG and PETRE-FDG images in the cycleGAN and pix2pix models were evaluated using the independent Student's t-test. Statistical significance was set at p ≤ 0.05. Results: The participant demographics (age, sex, or diagnosis) showed no statistically significant differences between the training (82 participants) and validation (28 participants) groups. The mean SSIM between the PETGE-FDG and PETRE-FDG images was 0.768 ± 0.135 for the cycleGAN model and 0.745 ± 0.143 for the pix2pix model. The mean PSNR was 32.4 ± 9.5 and 30.7 ± 8.0. The PETGE-FDG images of the cycleGAN model showed statistically higher mean SSIM than those of the pix2pix model (p < 0.001). The mean PSNR was also higher in the PETGE-FDG images of the cycleGAN model than those of pix2pix model (p < 0.001). Conclusions: We generated PETFDG images from PETFBB images using deep learning. The cycleGAN model generated PETGE-FDG images with a higher SSIM and PSNR values than the pix2pix model. Image-to-image translation using deep learning may be useful for generating PETFDG images. These may provide additional information for the management of Alzheimer's disease without extra image acquisition and the consequent increase in radiation exposure, inconvenience, or expenses.

PLANT DISEASE IDEDNTIFICATION USING MACHINE LEARNING AND IMAGE PROCESSING

The primary objective of this study is to investigate the detection and diagnosis of plant diseases using Deep Learning and Digital Image Processing. Previous research has primarily focused on single plant disease scenarios using publicly available datasets, often overlooking the image preprocessing phase. In this study, we propose a model that works with 10 different plants and utilizes approximately 50,000 images for training and testing. We classified 36 distinct classes into healthy or infected types based on disease labels. To enhance the accuracy of disease detection, we recommend employing image processing techniques and considering multiple plant scenarios. We utilized a dual-layer Convolutional Neural Network (CNN) for the publicly available dataset and supplemented it with real-time images captured from various farms in Village Rancharda Near Ahmedabad, Gujarat, India (PIN: 38255). Our research introduces several novel elements in the preprocessing steps. We employed HSV segmentation, flood fills segmentation, and a proposed deep learning model for image segmentation. Additionally, we standardized the resolution of all images to ensure uniformity. These preprocessing techniques refine the image data required for accurate classification and enhance the visibility of diseased portions. For image processing, we employed a sliding window mean average deviation technique and stacked the processed images onto the original image, resulting in six-channel images. Our proposed model demonstrates improved performance on the validation data, achieving an accuracy of up to 97.95%. Furthermore, we transformed this model into a TFLite model, which can be easily integrated into client applications without the need for a server. In our case, we implemented the model on an Android platform. These findings indicate the potential of our proposed model to significantly enhance the detection and diagnosis of plant diseases in real-world scenarios.

Knowledge Enhanced Semantic Communication Receiver

In recent years, with the rapid development of deep learning and natural language processing technologies, semantic communication has become a topic of great interest in the field of communication. Although existing deep learning-based semantic communication approaches have shown many advantages, they still do not make sufficient use of prior knowledge. Moreover, most existing semantic communication methods focus on the semantic encoding at the transmitter side, while we believe that the semantic decoding capability of the receiver should also be concerned. In this paper, we propose a knowledge enhanced semantic communication framework in which the receiver can more actively utilize the facts in the knowledge base for semantic reasoning and decoding, on the basis of only affecting the parameters rather than the structure of the neural networks at the transmitter side. Specifically, we design a transformer-based knowledge extractor to find relevant factual triples for the received noisy signal. Extensive simulation results on the WebNLG dataset demonstrate that the proposed receiver yields superior performance on top of the knowledge graph enhanced decoding.

Syntactic Versus Semantic Similarity of Artificial and Real Faults in Mutation Testing Studies

Fault seeding is typically used in empirical studies to evaluate and compare test techniques. Central to these techniques lies the hypothesis that artificially seeded faults involve some form of realistic properties and thus provide realistic experimental results. In an attempt to strengthen realism, a recent line of research uses machine learning techniques, such as deep learning and Natural Language Processing, to seed faults that look like (syntactically) real ones, implying that fault realism is related to syntactic similarity. This raises the question of whether seeding syntactically similar faults indeed results in semantically similar faults and, more generally whether syntactically dissimilar faults are far away (semantically) from the real ones. We answer this question by employing 4 state-of-the-art fault-seeding techniques (PiTest - a popular mutation testing tool, IBIR - a tool with manually crafted fault patterns, DeepMutation - a learning-based fault seeded framework and μBERT - a mutation testing tool based on the pre-trained language model CodeBERT) that operate in a fundamentally different way, and demonstrate that syntactic similarity does not reflect semantic similarity. We also show that 65.11%, 76.44%, 61.39% and 9.76% of the real faults of Defects4J V2 are semantically resembled by PiTest, IBIR, μBERT and DeepMutation faults, respectively.

Image Captioning Using Deep Learning

Abstract: This paper focuses on developing an image captioning system using deep learning techniques. The paper aims to generate descriptive textual captions for images, enabling machines to understand and communicate the content of visual data. The methodology involves leveraging convolutional neural networks (CNNs) for image feature extraction and recurrent neural networks (RNNs) for sequential language generation. The paper includes steps such as dataset collection, data preprocessing, CNN feature extraction, RNN-based captioning model implementation, model evaluation using metrics like BLEU score and METEOR, and presenting the results obtained. The expected deliverables include a functional image captioning system, comprehensive documentation, and a well-documented codebase. Through this paper, students gain practical experience in deep learning, computer vision, and natural language processing, contributing to advancements in image understanding and humanmachine interaction with visual data

Image-Processing-Based Subway Tunnel Crack Detection System.

With the increase in urban rail transit construction, instances of tunnel disease are on the rise, and cracks have become the focus of tunnel maintenance and management. Therefore, it is essential to carry out crack detection in a timely and efficient manner to not only prolong the service life of the tunnel but also reduce the incidence of accidents. In this paper, the design and structure of a tunnel crack detection system are analyzed. On this basis, this paper proposes a new method for crack identification and feature detection using image processing technology. This method fully considers the characteristics of tunnel images and the combination of these characteristics with deep learning, while a deep convolutional network (Single-Shot MultiBox Detector (SSD)) is proposed based on deep learning for object detection in complex images. The experimental results show that the test set accuracy and training set accuracy of the support vector machine (SVM) in the classification comparison test are up to 88% and 87.8%, respectively; while the test accuracy of Alexnet's deep convolutional neural network-based classification and identification is up to 96.7%, and the training set accuracy is up to 97.5%. It can be seen that this deep convolutional network recognition algorithm based on deep learning and image processing is better and more suitable for the detection of cracks in subway tunnels.

Fault Detection and Localization in Industrial IoT Systems using Deep Learning

In the ever-evolving landscape of Industrial Internet of Things (IoT) systems, the need for robust fault detection and localization has become paramount. Existing methods, while commendable, have faced certain limitations that necessitate the development of more sophisticated solutions. This work embarks on a journey to address these challenges and presents a novel approach to fault detection and localization, leveraging the formidable power of Deep Learning process. The limitations of existing methodologies primarily revolve around their inability to provide the required precision and accuracy in detecting and localizing faults within complex IoT deployments. These methods often struggle to handle the diverse contextual datasets that characterize modern industrial settings. Additionally, their computational efficiency often leaves much to be desired, hindering real-time fault response mechanisms. In response to these limitations, our proposed model introduces a hybrid framework. It combines Vector AutoRegressive Moving Average with Exogenous Variables (VARMAx) for precise fault localization and Convolutional Neural Networks (CNN) for robust fault detection. This fusion harnesses the strengths of both approaches, providing a comprehensive solution for fault management in industrial IoT systems. The advantages of this model are twofold. First, VARMAx offers exceptional accuracy in pinpointing the exact location of faults within the system, facilitating swift corrective actions. Second, the integration of CNN enhances fault detection by effectively capturing patterns and anomalies in the data, resulting in a highly responsive fault detection system. Notably, this model not only outperforms existing methods with a 3.9% boost in precision, 4.5% increase in accuracy, 4.9% higher recall, 8.5% improved AUC, but also boasts a remarkable 5.9% enhancement in computational speed, ensuring real-time responsiveness. The profound impact of this work extends to the realms of industrial automation and IoT system reliability. By addressing the limitations of existing approaches and introducing a robust fault detection and localization model, this paper paves the way for safer and more efficient industrial operations. The improved precision, accuracy, and speed of fault detection will minimize downtime, reduce maintenance costs, and ultimately elevate the performance and reliability of industrial IoT systems to new heights.

A deep learning-based approach for the automatic measurement of laser-cladding coating sizes

Problems such as long consumed time, high requirements for experience, and large subjectivity exist in the extraction of dimensional data of each area of the cross-section in the measurement process of the geometrical morphology of the laser cladding coatings. These problems seriously affect the accuracy of the coating morphology acquisition. An automatic measurement method for laser cladding coating sizes was studied based on deep learning and image processing technology. First, improved semantic segmentation model VMA-Unet of Unet deep learning was proposed. Each area of the coating morphology image was accurately divided, with the mIOU reaching 95.10 %. The image processing technology was then used to process predicted image output by the semantic segmentation model to automatically output the key coating size. We found that the mean accuracy of the measurement results of the method for each key size reached more than 98 % through tests and verification. A new method was provided for the accurate acquisition of the morphology data of the laser cladding coatings after verifying the feasibility and effectiveness of the automatic measurement method.

Infant Cry Signal Diagnostic System Using Deep Learning and Fused Features.

Early diagnosis of medical conditions in infants is crucial for ensuring timely and effective treatment. However, infants are unable to verbalize their symptoms, making it difficult for healthcare professionals to accurately diagnose their conditions. Crying is often the only way for infants to communicate their needs and discomfort. In this paper, we propose a medical diagnostic system for interpreting infants' cry audio signals (CAS) using a combination of different audio domain features and deep learning (DL) algorithms. The proposed system utilizes a dataset of labeled audio signals from infants with specific pathologies. The dataset includes two infant pathologies with high mortality rates, neonatal respiratory distress syndrome (RDS), sepsis, and crying. The system employed the harmonic ratio (HR) as a prosodic feature, the Gammatone frequency cepstral coefficients (GFCCs) as a cepstral feature, and image-based features through the spectrogram which are extracted using a convolution neural network (CNN) pretrained model and fused with the other features to benefit multiple domains in improving the classification rate and the accuracy of the model. The different combination of the fused features is then fed into multiple machine learning algorithms including random forest (RF), support vector machine (SVM), and deep neural network (DNN) models. The evaluation of the system using the accuracy, precision, recall, F1-score, confusion matrix, and receiver operating characteristic (ROC) curve, showed promising results for the early diagnosis of medical conditions in infants based on the crying signals only, where the system achieved the highest accuracy of 97.50% using the combination of the spectrogram, HR, and GFCC through the deep learning process. The finding demonstrated the importance of fusing different audio features, especially the spectrogram, through the learning process rather than a simple concatenation and the use of deep learning algorithms in extracting sparsely represented features that can be used later on in the classification problem, which improves the separation between different infants' pathologies. The results outperformed the published benchmark paper by improving the classification problem to be multiclassification (RDS, sepsis, and healthy), investigating a new type of feature, which is the spectrogram, using a new feature fusion technique, which is fusion, through the learning process using the deep learning model.

Unveiling the potential of pre-processing and differentiable binarization thresholds for invoice text detection

Invoice text detection aims to automatically identify key information in invoices, which is one of the representative applications of Optical Character Recognition (OCR) technology and a research hotspot in the computer vision community. The classic traditional OCR method consists of three key steps: preprocessing stage, text detection and text recognition. In the feature extraction stage, it faces problems such as insufficient generalization ability and poor robustness. In this paper, we reveal the great potential of preprocessing strategies in improving OCR accuracy, and propose an invoice text detection method based on DBNet and differentiable binarization thresholds. Specifically, we first introduce color segmentation and local adaptive threshold to improve the preprocessing process, which can effectively suppress the influence of background information and context noise on detection results. In addition, a differentiable binarization threshold is introduced in the feature extraction, which improves the error correction speed in the whole deep learning process. Studies have shown significant improvements in test results after color segmentation or background removal compared to the original image. Therefore, we propose to properly consider the text detection of invoices in the preprocessing stage for improvement.

Residual H-Net: A high-level feature extraction approach for brain lesion segmentation

In the case of acute stroke patients, mere diagnosis falls short, and segmentation is needed. Recent development in deep learning and image processing has provided us with the potential to automatically perform brain lesion segmentation. However, many of the approaches ended up failing to generalize to new data by overfitting the ATLAS R1.2 dataset and ignoring information extraction of the high-level features. We propose a novel Residual H-Net that addresses these two issues by adding a special residual block in the middle of the U-Net and increasing dilation size to better extract the high-level features. The presented model is less susceptible to overfitting and much easier to train. The Residual H-Net is tested on a subset of ATLAS R2.0 data and shows promising performance against the previous state-of-the-art model.

BERT‐TriF: An inductive short text classification model for power equipment defect records

AbstractThe descriptions of power equipment defect records are often characterized by colloquial short texts. Standardized classification of a large number of colloquial defect descriptions has laid a solid foundation for building a power equipment knowledge graph and improving the level of intelligence in the field of power inspection. Using deep learning and natural language processing technology, this paper proposes a text classification model for power equipment defect records named Bidirectional Encoder Representations from Transformers with Family Feature Fusion (BERT‐TriF). The model firstly leverages BERT to semantically represent the input text. To extract family history as well as text implicit information, we creatively propose a family feature fusion algorithm for training. An improved multi‐head attention mechanism is developed subsequently to enhance text semantic category features and strengthen the learning ability of the model. By comparing BERT‐TriF and baseline models such as TextCNN, TextRNN, and fastText on the specified and generic text dataset, the experimental results demonstrate that it has better performance, robustness, and universality for short text classification.

Crystal search - feasibility study of a real-time deep learning process for crystallization well images.

To avoid the time-consuming and often monotonous task of manual inspection of crystallization plates, a Python-based program to automatically detect crystals in crystallization wells employing deep learning techniques was developed. The program uses manually scored crystallization trials deposited in a database of an in-house crystallization robot as a training set. Since the success rate of such a system is able to catch up with manual inspection by trained persons, it will become an important tool for crystallographers working on biological samples. Four network architectures were compared and the SqueezeNet architecture performed best. In detecting crystals AlexNet accomplished a better result, but with a lower threshold the mean value for crystal detection was improved for SqueezeNet. Two assumptions were made about the imaging rate. With these two extremes it was found that an image processing rate of at least two times, but up to 58 times in the worst case, would be needed to reach the maximum imaging rate according to the deep learning network architecture employed for real-time classification. To avoid high workloads for the control computer of the CrystalMation system, the computing is distributed over several workstations, participating voluntarily, by the grid programming system from the Berkeley Open Infrastructure for Network Computing (BOINC). The outcome of the program is redistributed into the database as automatic real-time scores (ARTscore). These are immediately visible as colored frames around each crystallization well image of the inspection program. In addition, regions of droplets with the highest scoring probability found by the system are also available as images.

Revamping the Workout Routine : An Overview of PoseNet Thunder-Driven Fitness Apps Incorporating Computer Vision and Machine Learning

In today's modern lifestyle, characterised by a sedentary routine, the prevalence of chronic diseases among individuals has increased significantly. Juggling a 9-5 job, family responsibilities, and social commitments leaves little time for commuting to a gym regularly. As a result, more people are opting to exercise at home, leveraging the convenience of having fitness resources available through their handheld devices. This shift towards home workouts is facilitated by the remarkable advancements in technology. Using Different innovative approaches leverage the power of deep learning algorithms to analyse exercise routines and classify different types of workouts. Through computer vision, the system can process visual information from workout videos, enabling the recognition and understanding of various body movements, postures, and exercise techniques. This provides users with valuable insights and feedback on their performance, ensuring that they are executing exercises correctly and effectively. The integration of deep learning, computer vision, and image processing technologies offers immense potential in the field of exercise analysis and classification. Consistently performing an exercise improperly may result in significant long-term injury. We propose a method to assess the user's body posture during a workout and compare it to a professional's reference work out to assist address this problem and provide visual feedback while conducting a workout. To detect faults and deliver remedial action to the user, we model the human body as a collection of limbs and assess angles between limb pairs. Our system builds on the latest advancements using deep learning for human body pose estimation.

Techniques of deep learning and image processing in plant leaf disease detection: a review

&lt;span lang="EN-US"&gt;Computer vision techniques are an emerging trend today. Digital image processing is gaining popularity because of the significant upsurge in the usage of digital images over the internet. Digital image processing is a practice that can help in designing sophisticated high-end machines, which can hold &lt;span&gt;the ophthalmic functionality of the human eye. In agriculture, leaf examination&lt;/span&gt; is important for disease identification and fair warning for any deficiency within the plant. Many prominent plant species are facing extinction because of a lack of knowledge. A proper realization of computer vision techniques aid in extracting a significant amount of information from leaf image. This necessitates the requirement of an automatic leaf disease detection method to diagnose disease occurrences and severity, for timely crop management, by spraying pesticides. This study focuses on techniques of digital image processing and machine learning rendered in plant leaf disease detection, which has great potential in precision agriculture. To support this study, techniques exercised by various researchers in recent years are tabulated.&lt;/span&gt;

Image-based estimation of the left ventricular cavity volume using deep learning and Gaussian process with cardio-mechanical applications.

In this investigation, an image-based method has been developed to estimate the volume of the left ventricular cavity using cardiac magnetic resonance (CMR) imaging data. Deep learning and Gaussian processes have been applied to bring the estimations closer to the cavity volumes manually extracted. CMR data from 339 patients and healthy volunteers have been used to train a stepwise regression model that can estimate the volume of the left ventricular cavity at the beginning and end of diastole. We have decreased the root mean square error (RMSE) of cavity volume estimation approximately from 13 to 8ml compared to the common practice in the literature. Considering the RMSE of manual measurements is approximately 4ml on the same dataset, 8ml of error is notable for a fully automated estimation method, which needs no supervision or user-hours once it has been trained. Additionally, to demonstrate a clinically relevant application of automatically estimated volumes, we inferred the passive material properties of the myocardium given the volume estimates using a well-validated cardiac model. These material properties can be further used for patient treatment planning and diagnosis.

Rapid Analysis of Thorax Images for the Detection of Viral Infections

At the end of December 2019, a person in the Chinse city Wuhan was probably infected for the first time with the novel SARS-CoV-2 virus. In order to be able to react as quickly as possible after infection rapid diagnostic measures are of the utmost importance so that medical treatment can be taken at an early stage. An imageprocessing algorithm is presented using chest X-rays to determine whether a lung infection has a viral or a bacterial cause. In comparison to other more complicated evaluation methods, focus was put on using a simple algorithm by using the Canny algorithm for edge detection of infected areas of the lung tissue instead of complex deep learning processes. Main advantage here is that the method is portable to many different computer systems with little effort and does not need huge computing power. This should contribute to a faster diagnosis of SARS-CoV-2 virus-infection, especially in medically underdeveloped areas.