Advances In Deep Learning Research Articles

Cross-Granularity Infrared Image Segmentation Network for Nighttime Marine Observations

Infrared image segmentation in marine environments is crucial for enhancing nighttime observations and ensuring maritime safety. While recent advancements in deep learning have significantly improved segmentation accuracy, challenges remain due to nighttime marine scenes including low contrast and noise backgrounds. This paper introduces a cross-granularity infrared image segmentation network CGSegNet designed to address these challenges specifically for infrared images. The proposed method designs a hybrid feature framework with cross-granularity to enhance segmentation performance in complex water surface scenarios. To suppress feature semantic disparity against different feature granularity, we propose an adaptive multi-scale fusion module (AMF) that combines local granularity extraction with global context granularity. Additionally, incorporating a handcrafted histogram of oriented gradients (HOG) features, we designed a novel HOG feature fusion module to improve edge detection accuracy under low-contrast conditions. Comprehensive experiments conducted on the public infrared segmentation dataset demonstrate that our method outperforms state-of-the-art techniques, achieving superior segmentation results compared to professional infrared image segmentation methods. The results highlight the potential of our approach in facilitating accurate infrared image segmentation for nighttime marine observation, with implications for maritime safety and environmental monitoring.

Deep Learning Approach for Suspicious Activity Detection from Surveillance Video

The present generation seeks ways to live their lives free of fear. This often prompts the need for enhanced law enforcement and security systems. Suspicion activity recognition in surveillance footage has become an essential component in modern security systems, focused around the problem of recognizing inappropriate behaviors without too much human time. The advent of deep learning has made an impact in computer vision tasks within such a way the real-time detection by the evaluation of large amounts of video footage becomes realistic. In this context the paper presents a model in which deep learning, particularly Convolutional Neural Networks (CNNs), are utilized for the behavioral pattern recognition from security footage. In addition, it also improves the efficacy of detection and reduces the shortcomings associated with these systems where only the post-event analysis is performed manually or automatically. In this paper, we demonstrate a novel IP model that analyzes live video and performs normal-suspicious–threat triage in real-time. Results indicate the validity of the method in different environmental and illumination conditions. An important objective that this system is designed for is to change the way security is practiced and delivered in today’s intelligent systems providing actionable intelligence and the ability to prevent threats rather than just respond to them making this a high growth opportunity for marketing high-risk and sensitive areas.

A CNN-accelerated workflow for stochastic seismic property estimation

As one of the major tools in resolving the non-uniqueness challenge in subsurface interpretation and reservoir characterization, stochastic inversion from post- and pre-stack seismic data remains challenging, which not only requires heavy computational resources but also relies on intensive manual supervision. Inspired by the recent advances in deep learning particularly convolutional neural networks (CNNs) for interdisciplinary data integration, this study proposes a deep learning workflow that enables stochastic property estimation by efficiently integrating seismic images with sparse wells. It starts with sampling a set of property prior models (PPMs) from densely-measured properties at well locations and corrupting local seismic patterns with Gaussian noise. The core idea is to train a structure-guided CNN by mapping the contaminated seismic with the sampled PPMs while enforcing structural consistency to avoid overfitting in the presence of sparse wells. Finally, the baseline and uncertainty of target properties are estimated by running multiple realizations of the trained CNN. As demonstrated by three examples, with minimum efforts of CNN architecture customization according to data availability, the proposed workflow can accommodate various use cases, including rock acoustic/elastic property estimation from 3D post-/angle-stack seismic and soil geotechnical properties from 2D ultra-high-resolution seismic. In all examples, the machine predictions match seismic patterns well and are of high lateral consistency.

Speech Emotion Recognition Using Feedforward Neural Network

Speech Emotion Recognition (SER) has gained increasing attention due to its application in fields such as human-computer interaction, healthcare, customer service automation, and affective computing. This review focuses on the application of Deep Neural Networks (DNNs), specifically Feedforward Neural Networks (FNNs), in detecting and classifying emotions from speech signals. Recent advancements in deep learning have significantly enhanced the ability of machines to interpret emotional cues from auditory data. This paper discusses the motivation, objectives, and scope of employing DNN FNN architectures for SER. We also address the technical feasibility and potential of this approach for real-time applications. By reviewing existing literature, this study aims to provide insights into the current progress, challenges, and future prospects of SER systems.

A highly efficient tunnel lining crack detection model based on Mini-Unet.

The accurate detection of tunnel lining cracks and prompt identification of their primary causes are critical for maintaining tunnel availability. The advancement of deep learning, particularly in the domain of convolutional neural network (CNN) for image segmentation, has made tunnel lining crack detection more feasible. However, the CNN-based technique for tunnel lining crack detection commonly prioritizes increasing algorithmic complexity to enhance detection accuracy, posing a challenge in balancing the accuracy of detection and the efficiency of the algorithm. Motivated by the superior performance of Unet in image segmentation, this paper proposes a lightweight tunnel lining crack detection model named Mini-Unet, which refined the Unet architecture and utilized depthwise separable convolutions (DSConv) to replace some standard convolution layers. In the optimization of the proposed model parameters, applying a hybrid loss function that integrated dice loss and cross-entropy loss effectively tackled the imbalance between crack and background categories. Several models were set up to contrast with Mini-Unet and the experimental results were analyzed. Mini-Unet achieves a mean intersection over union (MIoU) of 60.76%, a mean precision of 84.18%, and a frame per second (FPS) of 5.635, respectively. Mini-Unet outperforms several mainstream models, enabling rapid detection while maintaining identified accuracy and facilitating the practical application of AI power for real-time tunnel lining crack detection.

Software Refactoring Network: An Improved Software Refactoring Prediction Framework Using Hybrid Networking‐Based Deep Learning Approach

ABSTRACTSoftware refactoring plays a vital role in maintaining and improving the quality of software systems. The software refactoring network aims to connect developers, researchers, and practitioners to share knowledge, best practices, and tools related to refactoring. However, the network faces various challenges, such as the complexity of software systems, the diversity of refactoring techniques, and the need for automated and intelligent solutions to assist developers in making refactoring decisions. By leveraging deep learning techniques, the software refactoring network can enhance the speed, accuracy, and relevance of refactoring suggestions, ultimately improving the overall quality and maintainability of software systems. So, in this paper, an advanced deep learning–based software refactoring framework is proposed. The suggested model performs three phases as (a) data collection, (b) feature extraction, and (c) prediction of software refactoring. Initially, the data is collected from ordinary datasets. Then, the collected data is fed to the feature extraction stage, where the source code, process, and ownership metrics of all refactored and non‐refactored data are retrieved for further processing. After that, the extracted features are predicted using Adaptive and Attentive Dilation Adopted Hybrid Network (AADHN) techniques, in which it is performed using Deep Temporal Context Networks (DTCN) with a Bidirectional Long‐Short Term Memory (Bi‐LSTM) model. Here, the parameters in the hybrid networking model are optimized with the help of Constant Integer Updated Golden Tortoise Beetle Optimizer (CIU‐GTBO) for improving the prediction process. Therefore, the accuracy of the developed algorithm has achieved for different datasets, whereas it shows the value of 96.41, 96.38, 96.38, 96.38, 96.41, 96.38, and 96.39 for antlr4, junit, mapdb, mcMMO, mct, oryx, and titan. Also, the precision of the developed model has shown the better performance of 96.38, 96.32, 96.37, 96.33, 96.35, 96.37, and 96.31 for the datasets like antlr4, junit, mapdb, mcMMO, mct, oryx, and titan.

Comparative Evaluation of Sentiment Analysis Methods: From Traditional Techniques to Advanced Deep Learning Models

Abstract. Sentiment evaluation plays a crucial role in deciphering public perception and consumer responses in today's digital landscape. This investigation offers a thorough assessment of diverse sentiment evaluation techniques, contrasting conventional machine learning methodologies with cutting-edge deep learning frameworks. In particular, the research scrutinizes the efficacy of Bidirectional Encoder Representations from Transformers (BERT)-derived architectures (BERT-Base and Robustly Optimized BERT Pretraining Approach (RoBERTa)), Convolutional Neural Networks (CNN), Long Short-Term Memory Networks (LSTM), Support Vector Machines (SVM), and Naive Bayes classifiers. The study gauges these approaches based on their precision, recall, F1-metric, overall accuracy, and computational efficiency using an extensive sentiment evaluation dataset. The results reveal that BERT-based models, particularly RoBERTa, achieve the highest accuracy (87.44%) and F1-score (0.8746), though they also require the longest training time (approximately 3 hours). CNN and LSTM models strike a balance between performance and efficiency, while traditional methods like SVM and Naive Bayes offer faster training and deployment with moderate accuracy. The insights gained from this study are valuable for both researchers and practitioners, highlighting the trade-offs between model performance, computational demands, and practical deployment considerations in sentiment analysis applications.

Digital profiling of gene expression from histology images with linearized attention.

Cancer is a heterogeneous disease requiring costly genetic profiling for better understanding and management. Recent advances in deep learning have enabled cost-effective predictions of genetic alterations from whole slide images (WSIs). While transformers have driven significant progress in non-medical domains, their application to WSIs lags behind due to high model complexity and limited dataset sizes. Here, we introduce SEQUOIA, a linearized transformer model that predicts cancer transcriptomic profiles from WSIs. SEQUOIA is developed using 7584 tumor samples across 16 cancer types, with its generalization capacity validated on two independent cohorts comprising 1368 tumors. Accurately predicted genes are associated with key cancer processes, including inflammatory response, cell cycles and metabolism. Further, we demonstrate the value of SEQUOIA in stratifying the risk of breast cancer recurrence and in resolving spatial gene expression at loco-regional levels. SEQUOIA hence deciphers clinically relevant information from WSIs, opening avenues for personalized cancer management.

Analysis of the impact of deep learning know-how and data in modelling neonatal EEG.

The performance gains achieved by deep learning models nowadays are mainly attributed to the usage of ever larger datasets. In this study, we present and contrast the performance gains that can be achieved via accessing larger high-quality datasets versus the gains that can be achieved from harnessing the latest deep learning architectural and training advances. Modelling neonatal EEG is particularly affected by the lack of publicly available large datasets. It is shown that greater performance gains can be achieved from harnessing the latest deep learning advances than using a larger training dataset when adopting AUC as a metric, whereas using AUC90 or AUC-PR as metrics greater performance gains are achieved from using a larger dataset than harnessing the latest deep learning advances. In all scenarios the best performance is obtained by combining both deep learning advances and larger datasets. A novel developed architecture is presented that outperforms the current state-of-the-art model for the task of neonatal seizure detection. A novel method to fine-tune the presented model towards site-specific settings based on pseudo labelling is also outlined. The code and the weights of the model are made publicly available for benchmarking future model performances for neonatal seizure detection.

DeepRSMA: a cross-fusion based deep learning method for RNA-small molecule binding affinity prediction.

RNA is implicated in numerous aberrant cellular functions and disease progressions, highlighting the crucial importance of RNA-targeted drugs. To accelerate the discovery of such drugs, it is essential to develop an effective computational method for predicting RNA-small molecule affinity (RSMA). Recently, deep learning based computational methods have been promising due to their powerful nonlinear modeling ability. However, the leveraging of advanced deep learning methods to mine the diverse information of RNAs, small molecules and their interaction still remains a great challenge. In this study, we present DeepRSMA, an innovative cross-attention-based deep learning method for RSMA prediction. To effectively capture fine-grained features from RNA and small molecules, we developed nucleotide-level and atomic-level feature extraction modules for RNA and small molecules, respectively. Additionally, we incorporated both sequence and graph views into these modules to capture features from multiple perspectives. Moreover, a Transformer-based cross-fusion module is introduced to learn the general patterns of interactions between RNAs and small molecules. To achieve effective RSMA prediction, we integrated the RNA and small molecule representations from the feature extraction and cross-fusion modules. Our results show that DeepRSMA outperforms baseline methods in multiple test settings. The interpretability analysis and the case study on spinal muscular atrophy (SMA) demonstrate that DeepRSMA has the potential to guide RNA-targeted drug design. The codes and data are publicly available at https://github.com/Hhhzj-7/DeepRSMA. Supplementary data are available at Bioinformatics online.

A novel optimization-driven deep learning framework for the detection of DDoS attacks.

Distributed denial of service (DDoS) attack is one of the most hazardous assaults in cloud computing or networking. By depleting resources, this attack renders the services unavailable to end users and leads to significant financial and reputational damage. Hence, identifying such threats is crucial to minimize revenue loss, market share, and productivity loss and enhance the brand reputation. In this study, we implemented an effective intrusion detection system using deep learning approach. The suggested framework includes three phases: Data pre-processing, Data balancing, and Classification. First, we prepare the valid data, which is helpful for further processing. Then, we balance the given pre-processed data by Conditional generative adversarial network (CGAN), and as a result, we can minimize the bias towards the majority classes. Finally, we distinguish whether the traffic is attack or benign using a stacked sparse denoising autoencoder (SSDAE) with a firefly-black widow (FA-BW) hybrid optimization algorithm. All these experiments are validated through the CICDDoS2019 dataset and compared with well-received techniques. From these findings, we observed that the proposed strategy detects DDoS attacks significantly more accurately than other approaches. Based on our findings, this study highlights the crucial role played by advanced deep learning techniques and hybrid optimization algorithms in strengthening cybersecurity against DDoS attacks.

Technical Note: Neural Network Architectures for Self-Supervised Body Part Regression Models with Automated Localized Segmentation Application.

The advancement of medical image deep learning necessitates tools that can accurately identify body regions from whole-body scans to serve as an essential pre-processing step for downstream tasks. Typically, these deep learning models rely on labeled data and supervised learning, which is labor-intensive. However, the emergence of self-supervised learning is revolutionizing the field by eliminating the need for labels. The purpose of this study was to compare neural network architectures of self-supervised models that produced a body part regression (BPR) slice score to aid in the development of anatomically localized segmentation models. VGG, ResNet, DenseNet, ConvNext, and EfficientNet BPR models were implemented in the MONAI/Pytorch framework. Landmark organs were correlated to slice scores and mean absolute error (MAE) was calculated from the predicted slice and the actual slice of various organ landmarks. Four localized DynUNet segmentation models (thorax, upper abdomen, lower abdomen, and pelvis) were developed using the BPR slice scores. Dice similarity coefficient (DSC) was compared between the localized and baseline segmentation models. The best performing BPR model was the EfficientNet architecture with an overall 3.18 MAE, compared to the VGG baseline model with a MAE of 6.29. The localized segmentation model significantly outperformed the baseline in 16 out of 20 organs with a DSC of 0.88. Enhanced neural networks like EfficientNet have a large performance increase in localizing anatomical structures in a CT compared in BPR task. Utilizing BPR slice score is shown to be effective in anatomically localized segmentation tasks with improved performance.

EDH-STNet: An Evaporation Duct Height Spatiotemporal Prediction Model Based on Swin-Unet Integrating Multiple Environmental Information Sources

Given the significant spatial non-uniformity of marine evaporation ducts, accurately predicting the regional distribution of evaporation duct height (EDH) is crucial for ensuring the stable operation of radio systems. While machine-learning-based EDH prediction models have been extensively developed, they fail to provide the EDH distribution over large-scale regions in practical applications. To address this limitation, we have developed a novel spatiotemporal prediction model for EDH that integrates multiple environmental information sources, termed the EDH Spatiotemporal Network (EDH-STNet). This model is based on the Swin-Unet architecture, employing an Encoder–Decoder framework that utilizes consecutive Swin-Transformers. This design effectively captures complex spatial correlations and temporal characteristics. The EDH-STNet model also incorporates nonlinear relationships between various hydrometeorological parameters (HMPs) and EDH. In contrast to existing models, it introduces multiple HMPs to enhance these relationships. By adopting a data-driven approach that integrates these HMPs as prior information, the accuracy and reliability of spatiotemporal predictions are significantly improved. Comprehensive testing and evaluation demonstrate that the EDH-STNet model, which merges an advanced deep learning algorithm with multiple HMPs, yields accurate predictions of EDH for both immediate and future timeframes. This development offers a novel solution to ensure the stable operation of radio systems.

Dynamic inferences of crash risks in freeway merging zones: a spatio-temporal deep learning model

Freeway merging zones are critical for freeway operations and management due to potential crashes arising from complex vehicle merging behaviours. This paper investigates the application of a spatio-temporal deep learning model to infer crash risks in the zones. We first introduce a crash risk index based on Time-To-Collision and vehicle merging patterns. An innovation is that the developed spatio-temporal transformer model can analyze the evolving risk index. This model effectively captures dynamic risk features through a multi-head attention mechanism within its spatio-temporal learning components. Numerical experiments on nine inference tasks with varying spatial resolutions show improved performance of the model with lower resolution. Moreover, the ST-Transformer model is benchmarked against three advanced deep learning models, which consistently demonstrates its superiority in capturing spatio-temporal dependence in risk sequences. This investigation significantly contributes to a richer understanding of proactive traffic safety, providing valuable insights for advanced freeway management and driver assistance systems.

Mixed Reality-Based Concrete Crack Detection and Skeleton Extraction Using Deep Learning and Image Processing

Advancements in image processing and deep learning offer considerable opportunities for automated defect assessment in civil structures. However, these systems cannot work interactively with human inspectors. Mixed reality (MR) can be adopted to address this by involving inspectors in various stages of the assessment process. This paper integrates You Only Look Once (YOLO) v5n and YOLO v5m with the Canny algorithm for real-time concrete crack detection and skeleton extraction with a Microsoft HoloLens 2 MR device. The YOLO v5n demonstrates a superior mean average precision (mAP) 0.5 and speed, while YOLO v5m achieves the highest mAP 0.5 0.95 among the other YOLO v5 structures. The Canny algorithm also outperforms the Sobel and Prewitt edge detectors with the highest F1 score. The developed MR-based system could not only be employed for real-time defect assessment but also be utilized for the automatic recording of the location and other specifications of the cracks for further analysis and future re-inspections.

Predictive Health Monitoring of Palm Trees: Techniques and Applications for Enhanced Processing

This paper presents new framework for predictive health monitoring in palm trees using advanced deep learning methods. The system then selects the YOLOv8 method for the detection of the health of palm trees and further divides them into their real-time classification. Here, the model will, after training on its dataset, classify the palm tree as healthy, diseased, and stressed trees, keeping all the influential factors of the growth of the palm tree in consideration. It makes provision for the profiling of health status by way of early intervention, reducing risks and preventing diseases with a view to optimizing crop yields within palm plantations. The proposed model leverages state-of-the-art methodologies in object detection to process images of the palm tree with identification of key indicators of health issues. This approach has great implications for agricultural productivity in view of the maintenance of plant health through early detection. Further, the paper discusses challenges presented during the training and validation of the model, their strategies for overcoming such obstacles, and goes further into relating details to the technical architecture of the model. Further details on the works related to plant disease detection, image-based classification, and further applications of deep learning in agriculture have inspired the research work. Based on the obtained results, it can be observed that the proposed system provides a scalable and efficient solution for continuous monitoring and assessment of the health conditions of palm trees, hence enabling sustainable agricultural practices.

Automatic Classification of Focal Liver Lesions Based on Multi-Sequence MRI.

Accurate and automated diagnosis of focal liver lesions is critical for effective radiological practice and patient treatment planning. This study presents a deep learning model specifically developed for classifying focal liver lesions across eight different MRI sequences, categorizing them into seven distinct classes. The model includes a feature extraction module that derives multi-level representations of the lesions, a feature fusion attention module to integrate contextual information from the various sequences, and an attention-guided data augmentation module to enrich the training dataset. The proposed model achieved a patient-wise classification accuracy of 0.9302 and a lesion-wise accuracy of 0.8592, along with an F1-score of 0.8395, a recall of 0.8296, and a precision of 0.8551. These findings demonstrate the effectiveness of combining multi-sequence MRI with advanced deep learning methodologies, providing a robust tool to support radiologists in accurately classifying liver lesions in clinical settings.

Review of Recent Advances in Remote Sensing and Machine Learning Methods for Lake Water Quality Management

This review examines the integration of remote sensing technologies and machine learning models for efficient monitoring and management of lake water quality. It critically evaluates the performance of various satellite platforms, including Landsat, Sentinel-2, MODIS, RapidEye, and Hyperion, in assessing key water quality parameters including chlorophyll-a (Chl-a), turbidity, and colored dissolved organic matter (CDOM). This review highlights the specific advantages of each satellite platform, considering factors like spatial and temporal resolution, spectral coverage, and the suitability of these platforms for different lake sizes and characteristics. In addition to remote sensing platforms, this paper explores the application of a wide range of machine learning models, from traditional linear and tree-based methods to more advanced deep learning techniques like convolutional neural networks (CNNs), recurrent neural networks (RNNs), and generative adversarial networks (GANs). These models are analyzed for their ability to handle the complexities inherent in remote sensing data, including high dimensionality, non-linear relationships, and the integration of multispectral and hyperspectral data. This review also discusses the effectiveness of these models in predicting various water quality parameters, offering insights into the most appropriate model–satellite combinations for different monitoring scenarios. Moreover, this paper identifies and discusses the key challenges associated with data quality, model interpretability, and integrating remote sensing imagery with machine learning models. It emphasizes the need for advancements in data fusion techniques, improved model generalizability, and the developing robust frameworks for integrating multi-source data. This review concludes by offering targeted recommendations for future research, highlighting the potential of interdisciplinary collaborations to enhance the application of these technologies in sustainable lake water quality management.

Personalized Emotion Detection Adapting Models to Individual Emotional Expressions

Emotion recognition from text and speech has become a critical area of research in artificial intelligence (AI), enhancing human-computer interaction across various sectors. This paper explores the methodologies used in emotion recognition, focusing on Natural Language Processing (NLP) for text and acoustic analysis for speech. It reviews key machine learning and deep learning models, including Support Vector Machines (SVM), neural networks, and transformers, and highlights the datasets commonly used in emotion detection studies. The paper also addresses challenges such as multimodal integration, data ambiguity, and ethical considerations like privacy concerns and bias in models. Applications in customer service, healthcare, education, and entertainment are discussed, showcasing the growing importance of emotion recognition in AI- driven systems. Future research directions, including advancements in deep learning, multimodal systems, and real-time processing, are also explored to address existing limitations.

Deep Learning-Based Classification of Macrofungi: Comparative Analysis of Advanced Models for Accurate Fungi Identification

This study focuses on the classification of six different macrofungi species using advanced deep learning techniques. Fungi species, such as Amanita pantherina, Boletus edulis, Cantharellus cibarius, Lactarius deliciosus, Pleurotus ostreatus and Tricholoma terreum were chosen based on their ecological importance and distinct morphological characteristics. The research employed 5 different machine learning techniques and 12 deep learning models, including DenseNet121, MobileNetV2, ConvNeXt, EfficientNet, and swin transformers, to evaluate their performance in identifying fungi from images. The DenseNet121 model demonstrated the highest accuracy (92%) and AUC score (95%), making it the most effective in distinguishing between species. The study also revealed that transformer-based models, particularly the swin transformer, were less effective, suggesting room for improvement in their application to this task. Further advancements in macrofungi classification could be achieved by expanding datasets, incorporating additional data types such as biochemical, electron microscopy, and RNA/DNA sequences, and using ensemble methods to enhance model performance. The findings contribute valuable insights into both the use of deep learning for biodiversity research and the ecological conservation of macrofungi species.