Feature Extraction Model Research Articles

DRL-enhanced 3D detection of occluded stems for robotic grape harvesting

Detecting grape stems is essential for the autonomous operation of grape-picking robots. In natural orchards, the complex and irregular positioning of grape bunches, along with frequent occlusions caused by obstacles and leaves, create a dynamic and unpredictable environment that substantially affects the robot’s perception quality and harvesting efficiency. Inspired by human active observation mechanisms, this study introduces an end-to-end active visual perception framework using deep reinforcement learning (DRL) to enhance the detection of grape stems under complex occlusion. An instance segmentation network is trained to obtain 2D masks, which are integrated into an octree volumetric grid to produce detailed volumetric observations for DRL training. Different occupancy weights are assigned to both explored and newly discovered regions, contributing to a novel reward function based on information gain, which subsequently drives the network to optimize camera movements, guiding the robotic arm towards the best viewpoint for efficient grape stem detection. The superior performance of the proposed method has been validated in both laboratory and outdoor orchards settings. The primary contribution of this work lies in presenting a novel fully end-to-end detection framework for occluded grape stems. Compared to existing methods, it enables the system to directly learn optimal viewpoint strategies through interactions between the robotic arm and the environment, effectively handling feature extraction and environmental modeling without the need for manually designed information gain metrics. This advancement provides foundational support for the next generation of highly autonomous picking robots, capable of operating adaptively in complex, unstructured environments.

Detection of community emotions through Sound: An Investigation using the FF-Orbital Chaos-Based feature extraction model

Detection of community emotions through Sound: An Investigation using the FF-Orbital Chaos-Based feature extraction model

CIMIL-CRC: A clinically-informed multiple instance learning framework for patient-level colorectal cancer molecular subtypes classification from H&E stained images.

Treatment approaches for colorectal cancer (CRC) are highly dependent on the molecular subtype, as immunotherapy has shown efficacy in cases with microsatellite instability (MSI) but is ineffective for the microsatellite stable (MSS) subtype. There is promising potential in utilizing deep neural networks (DNNs) to automate the differentiation of CRC subtypes by analyzing hematoxylin and eosin (H&E) stained whole-slide images (WSIs). Due to the extensive size of WSIs, multiple instance learning (MIL) techniques are typically explored. However, existing MIL methods focus on identifying the most representative image patches for classification, which may result in the loss of critical information. Additionally, these methods often overlook clinically relevant information, like the tendency for MSI class tumors to predominantly occur on the proximal (right side) colon. We introduce 'CIMIL-CRC', a DNN framework that: (1) solves the MSI/MSS MIL problem by efficiently combining a pre-trained feature extraction model with principal component analysis (PCA) to aggregate information from all patches, and (2) integrates clinical priors, particularly the tumor location within the colon, into the model to enhance patient-level classification accuracy. We assessed our CIMIL-CRC method using the average area under the receiver operating characteristic curve (AUROC) from a 5-fold cross-validation experimental setup for model development on the TCGA-CRC-DX cohort, contrasting it with a baseline patch-level classification, a MIL-only approach, and a clinically-informed patch-level classification approach. Our CIMIL-CRC outperformed all methods (AUROC: 0.92±0.002 (95% CI 0.91-0.92), vs. 0.79±0.02 (95% CI 0.76-0.82), 0.86±0.01 (95% CI 0.85-0.88), and 0.87±0.01 (95% CI 0.86-0.88), respectively). The improvement was statistically significant. To the best of our knowledge, this is the best result achieved for MSI/MSS classification on this dataset. Our CIMIL-CRC method holds promise for offering insights into the key representations of histopathological images and suggests a straightforward implementation.

Superpixel guided spectral-spatial feature extraction and weighted feature fusion for hyperspectral image classification with limited training samples

Deep learning is a double-edged sword. The powerful feature learning ability of deep models can effectively improve classification accuracy. Still, when the training samples for each class are limited, it will not only face the problem of overfitting but also significantly affect the classification result. Aiming at this critical problem, we propose a novel model of spectral-spatial feature extraction and weighted fusion guided by superpixels. It aims to thoroughly “squeeze” and utilize the untapped spectral and spatial features contained in hyperspectral images from multiple angles and stages. Firstly, with the guidance of superpixels, we represent the hyperspectral image in the form of latent features and use the multi-band priority criterion to select the final discriminant features. Secondly, we design a pixel-based CNN and a two-scale superpixel-based GCN classification framework for weighted feature fusion. Compared with several excellent band selection methods, the superb performance of our feature extraction module is verified. In addition, under the condition of only five training samples for each class, we conducted comparative experiments with several of the state-of-the-art classification methods and verified the excellent performance of our method on three widely used data sets.

MammoViT: A Custom Vision Transformer Architecture for Accurate BIRADS Classification in Mammogram Analysis

Background: Breast cancer screening through mammography interpretation is crucial for early detection and improved patient outcomes. However, the manual classification of mammograms using the BIRADS (Breast Imaging-Reporting and Data System) remains challenging due to subtle imaging features, inter-reader variability, and increasing radiologist workload. Traditional computer-aided detection systems often struggle with complex feature extraction and contextual understanding of mammographic abnormalities. To address these limitations, this study proposes MammoViT, a novel hybrid deep learning framework that leverages both ResNet50’s hierarchical feature extraction capabilities and Vision Transformer’s ability to capture long-range dependencies in images. Methods: We implemented a multi-stage approach utilizing a pre-trained ResNet50 model for initial feature extraction from mammogram images. To address the significant class imbalance in our four-class BIRADS dataset, we applied SMOTE (Synthetic Minority Over-sampling Technique) to generate synthetic samples for minority classes. The extracted feature arrays were transformed into non-overlapping patches with positional encodings for Vision Transformer processing. The Vision Transformer employs multi-head self-attention mechanisms to capture both local and global relationships between image patches, with each attention head learning different aspects of spatial dependencies. The model was optimized using Keras Tuner and trained using 5-fold cross-validation with early stopping to prevent overfitting. Results: MammoViT achieved 97.4% accuracy in classifying mammogram images across different BIRADS categories. The model’s effectiveness was validated through comprehensive evaluation metrics, including a classification report, confusion matrix, probability distribution, and comparison with existing studies. Conclusions: MammoViT effectively combines ResNet50 and Vision Transformer architectures while addressing the challenge of imbalanced medical imaging datasets. The high accuracy and robust performance demonstrate its potential as a reliable tool for supporting clinical decision-making in breast cancer screening.

Deep learning in microbiome analysis: a comprehensive review of neural network models

Microbiome research, the study of microbial communities in diverse environments, has seen significant advances due to the integration of deep learning (DL) methods. These computational techniques have become essential for addressing the inherent complexity and high-dimensionality of microbiome data, which consist of different types of omics datasets. Deep learning algorithms have shown remarkable capabilities in pattern recognition, feature extraction, and predictive modeling, enabling researchers to uncover hidden relationships within microbial ecosystems. By automating the detection of functional genes, microbial interactions, and host-microbiome dynamics, DL methods offer unprecedented precision in understanding microbiome composition and its impact on health, disease, and the environment. However, despite their potential, deep learning approaches face significant challenges in microbiome research. Additionally, the biological variability in microbiome datasets requires tailored approaches to ensure robust and generalizable outcomes. As microbiome research continues to generate vast and complex datasets, addressing these challenges will be crucial for advancing microbiological insights and translating them into practical applications with DL. This review provides an overview of different deep learning models in microbiome research, discussing their strengths, practical uses, and implications for future studies. We examine how these models are being applied to solve key problems and highlight potential pathways to overcome current limitations, emphasizing the transformative impact DL could have on the field moving forward.

Predicting EV battery state of health using long short term degradation feature extraction and FEA TimeMixer

Accurately predicting the State of Health (SOH) of new energy vehicle batteries is critical for ensuring their reliable operation and extending battery’s service life. To address the issue of low SOH prediction accuracy across different prediction lengths, this paper proposes a prediction method based on long-short-term battery degradation feature extraction and FEA-TimeMixer model. First, a novel automatic SOH extraction algorithm for offline charging data is introduced to label the battery SOH degradation data. Then, the autoencoder is utilized to fuse the features of long-term and short-term SOH degradation trends extracted by empirical degradation models and Complete Ensemble Empirical Mode Decomposition with Adaptive Noise to improve the prediction accuracy over different prediction lengths. Finally, a Frequency Enhanced Attention (FEA) mechanism is introduced to improve the TimeMixer model, which integrates time-domain and frequency-domain information to overcome the limitations of the original model in capturing frequency-domain features. Experimental results show that the proposed method achieves a Mean Absolute Error of less than 0.0219 for short-term SOH predictions and less than 0.1007 for long-term SOH predictions, outperforming other deep learning models in prediction accuracy over multiple prediction lengths.

Deep-AMPpred: A Deep Learning Method for Identifying Antimicrobial Peptides and Their Functional Activities.

Antimicrobial peptides (AMPs) are small peptides that play an important role in disease defense. As the problem of pathogen resistance caused by the misuse of antibiotics intensifies, the identification of AMPs as alternatives to antibiotics has become a hot topic. Accurately identifying AMPs using computational methods has been a key issue in the field of bioinformatics in recent years. Although there are many machine learning-based AMP identification tools, most of them do not focus on or only focus on a few functional activities. Predicting the multiple activities of antimicrobial peptides can help discover candidate peptides with broad-spectrum antimicrobial ability. We propose a two-stage AMP predictor deep-AMPpred, in which the first stage distinguishes AMP from other peptides, and the second stage solves the multilabel problem of 13 common functional activities of AMP. deep-AMPpred combines the ESM-2 model to encode the features of AMP and integrates CNN, BiLSTM, and CBAM models to discover AMP and its functional activities. The ESM-2 model captures the global contextual features of the peptide sequence, while CNN, BiLSTM, and CBAM combine local feature extraction, long-term and short-term dependency modeling, and attention mechanisms to improve the performance of deep-AMPpred in AMP and its function prediction. Experimental results demonstrate that deep-AMPpred performs well in accurately identifying AMPs and predicting their functional activities. This confirms the effectiveness of using the ESM-2 model to capture meaningful peptide sequence features and integrating multiple deep learning models for AMP identification and activity prediction.

Pendekatan Hybrid Untuk Deteksi Tepi Citra Menggunakan Algoritma Pembelajaran Mendalam DAM Pemrosesan Berbasis Morfologi

In the digital era, image edge detection plays an important role in various image processing applications such as medical analysis and object recognition. This research develops a hybrid model that combines the high accuracy of deep learning with the efficiency and robustness to noise of morphological processing to improve edge detection performance. The method used involves deep learning models for image feature extraction and morphological processing techniques to improve detection results. The research results show that this hybrid model is able to overcome the weaknesses of traditional methods and pure deep learning by producing edge detection that is smoother and more continuous, and more efficient in processing time. Tests on the BSDS500 dataset and medical images show that this model provides consistent performance in a variety of image conditions, including images with high noise and low contrast. This hybrid approach offers a practical and effective solution for a variety of image processing applications.

A comparative study of pre-trained models for image feature extraction in weather image classification using orange data mining

This paper presents a detailed comparative analysis of pre-trained models for feature extraction in the domain of weather image classification. Utilizing the orange data mining toolkit, we investigated the effectiveness of six prominent pre-trained models-InceptionV3, SqueezeNet, VGG-16, VGG-19, painter, and DeepLoc-in accurately classifying weather phenomena images. Among these models, InceptionV3, in conjunction with neural networks, emerged as the most effective, achieving a classification accuracy (CA) of 96.1%. Painter and SqueezeNet also showed strong performance, with accuracies of 95.1% and 86.7%, respectively, although they were surpassed by InceptionV3. VGG-16 and VGG-19 provided moderate accuracy, while DeepLoc underperformed significantly with a maximum accuracy of 56%. Neural networks consistently outperformed other classifiers across all models. This study highlights the critical importance of selecting appropriate pre-trained models to enhance the accuracy and reliability of weather image classification systems.

Artificial Intelligence based Automated Sign Gesture Recognition Solutions for Visually Challenged People

Gesture recognition is employed in human-machine communications, enhancing human life with impairments or who depend on non-verbal instructions. Hand gestures role an important role in the field of assistive technology for persons with visual impairments, whereas an optimum user communication design is of major importance. Many authors with substantial development for gesture recognition modeled several methods by using deep learning (DL) methods. This article introduces a Robust Gesture Sign Language Recognition Using Chicken Earthworm Optimization with Deep Learning (RSLR-CEWODL) approach. The projected RSLR-CEWODL algorithm majorly focuses on the recognition and classification of sign language. To accomplish this, the presented RSLR-CEWODL technique utilizes a residual network (ResNet-101) model for feature extraction. For optimal hyper parameter tuning process, the presented RSLR-CEWODL algorithm exploits the CEWO algorithm. Besides, the RSLR-CEWODL technique uses a whale optimization algorithm (WOA) with deep belief network (DBN) method for the sign language recognition method. The simulation result of the RSLR-CEWODL algorithm is tested using sign language datasets and the outcome was measured under various measures. The simulation values demonstrated the enhancements of the RSLR-CEWODL technique over other methodologies.

Multiangle correlation feature extraction and disease prediction model construction for patients with post-stroke dysarthria

Multiangle correlation feature extraction and disease prediction model construction for patients with post-stroke dysarthria

Deep learning-driven UAV vision for automated road crack detection and classification

ABSTRACT Unmanned aerial vehicles are increasingly utilised for monitoring and inspecting critical infrastructure such as power generation grids, oil and gas pipelines and roads. One key task in road maintenance is the detection of cracks, which is crucial for ensuring road safety. Manual detection of cracks is time-consuming and prone to errors, highlighting the need for automated solutions. This study presents an automated method for road crack detection and classification using a hybrid deep learning technique. The proposed approach integrates a pyramid vision transformer and ConvMixer models for feature extraction, enabling the system to learn complex patterns in crack images. Image pre-processing is first performed using a median filtering technique to enhance image quality. The detection and classification of cracks are then carried out using an Elman neural network model, with its hyperparameters optimised through an improved black widow optimisation algorithm. Extensive simulations demonstrate that the proposed method outperforms other deep learning models in terms of performance, providing a reliable and efficient solution for automated crack detection.

CT-RURnet: a novel network design for radar unmanned aerial vehicles recognition

Abstract The rapid growth of commercial unmanned aerial vehicles (UAVs) has led to more incidents. These include illegal activities by malicious actors. Such activities pose significant threats to public facilities and people’s safety. Therefore, distinguishing UAVs from other targets is a critical measure in preventing potential hazards. This paper proposes an enhanced feature extraction and global modeling model for UAV recognition, named CT-RURnet. The model directly learns from radar echo data, firstly using the convolutional neural networks module to quickly extract local features, and then using the Transformer module to establish global connections among these local features. The proposed model primarily focuses on learning the micro-Doppler features produced by targets, which facilitates UAV recognition. The used datasets include both simulated data and the Real Doppler RAD-DAR (RDRD) dataset. The proposed network achieves accuracy rates of 98.7% (10 SNR), 97.9% (5 SNR), and 86.95% (0 SNR) on the simulated dataset under varying SNR conditions and 97.14% on the real dataset. Compared to other baseline models, the CT-RURnet consistently delivers superior results. Ablation experiments are conducted to verify the contribution of each module to the overall network performance.

Semantic-Aware Hybrid Deep Learning Model for Brain Tumor Detection and Classification Using Adaptive Feature Extraction and Mask-RCNN

A brain tumor is one of the most prevalent causes of cancer death. The best strategy is the timely treatment of brain tumors in their early detection. Magnetic Resonance Imaging (MRI) is a standard non-invasive method to detect brain tumors. For early detection and better patient survival through MRI scans, the diagnosis needs a high level of knowledge in the radiological and neurological domains to identify the cancers. Researchers have suggested various brain cancer detection techniques. However, most existing automatic cancer detection approaches suffer from poor accuracy and low detection rates. This paper proposes a hybrid deep learning (DL) using deep feature extraction and adaptive Mask Region-based Convolutional Neural Networks (Mask-RCNNs) model for brain tumor detection and classification method to overcome these issues. The experimental findings on the benchmark dataset demonstrate that the planned model is highly effective, with 99.64% accuracy, 95.93% precision, 95.39% recall, and 95.67% F1-score.

Brand Logos Recognition System Using Image Processing for Food and Beverage Brands

This study investigates the development of a Brand Logo Recognition (BLR) system employing Convolutional Neural Networks (CNNs), specifically designed for the food and beverage industry in Ipoh. Accurate logo recognition is vital for businesses to strengthen brand identity, monitor consumer engagement, and mitigate the misuse of counterfeit logos. Existing systems often encounter challenges related to variations in logo design, image quality, and lighting conditions. To address these issues, the research adopts a hybrid methodology that integrates the Machine Learning Life Cycle and the Software Development Life Cycle (SDLC), utilizing an iterative Agile development framework. The system incorporates CNN models for feature extraction and classification, complemented by Single Shot Detector (SSD) algorithms for object detection. A curated dataset of food and beverage logos underwent preprocessing techniques, including resizing, normalization, and augmentation, to enhance the model’s generalization capabilities. Empirical results demonstrate high accuracy in detecting and classifying logos across diverse conditions, underscoring the effectiveness of the CNN-SSD architecture. The proposed system offers practical applications for marketing analytics and consumer research, empowering local businesses to refine branding strategies and improve customer engagement. Future research directions include the exploration of multi-label classification, real-time processing, and the integration of advanced methodologies, such as generative adversarial networks (GANs), for counterfeit logo detection. This study emphasizes the transformative potential of AI-driven logo recognition systems in revolutionizing marketing practices and supporting small and medium-sized enterprises (SMEs).

YOLO‐Tiny: A lightweight small object detection algorithm for UAV aerial imagery

AbstractIn unmanned aerial vehicle (UAV) aerial target detection tasks, two main challenges exist: the limited computational resources of UAV terminals, which are not conducive to running complex models, and the prevalence of small targets, which can easily lead to missed detections and false positives. To address these issues, this study proposes a lightweight and high‐accuracy small‐object detection algorithm for UAV aerial imagery that is based on YOLOv5s. First, the network structure is optimized by removing the layers in YOLOv5s primarily used for detecting large targets (P4 and P5) and adding layers primarily used for detecting small targets (P2 and P3). This enables the model to focus more on extracting small‐object features. A lightweight dynamic convolution is subsequently introduced in the C3 module, and the lightweight LW_C3 and LW_downsampling modules are designed for feature extraction and downsampling operations. This enhances the model's feature extraction capability while achieving a lightweight design. Finally, the adaptive multi‐scale spatial feature fusion (AMSFF) module is designed to adaptively learn the spatial weights of the feature maps at different levels, thereby further strengthening the effective fusion of multi‐scale features. Experimental results show that the improved YOLO‐Tiny model has higher accuracy and lower complexity, hence validating its excellent performance.

Deep Learning-Based Diagnosis Algorithm for Alzheimer's Disease.

Alzheimer's disease (AD), a degenerative condition affecting the central nervous system, has witnessed a notable rise in prevalence along with the increasing aging population. In recent years, the integration of cutting-edge medical imaging technologies with forefront theories in artificial intelligence has dramatically enhanced the efficiency of identifying and diagnosing brain diseases such as AD. This paper presents an innovative two-stage automatic auxiliary diagnosis algorithm for AD, based on an improved 3D DenseNet segmentation model and an improved MobileNetV3 classification model applied to brain MR images. In the segmentation network, the backbone network was simplified, the activation function and loss function were replaced, and the 3D GAM attention mechanism was introduced. In the classification network, firstly, the CA attention mechanism was added to enhance the model's ability to capture positional information of disease features; secondly, dilated convolutions were introduced to extract richer features from the input feature maps; and finally, the fully connected layer of MobileNetV3 was modified and the idea of transfer learning was adopted to improve the model's feature extraction capability. The results of the study showed that the proposed approach achieved classification accuracies of 97.85% for AD/NC, 95.31% for MCI/NC, 93.96% for AD/MCI, and 92.63% for AD/MCI/NC, respectively, which were 3.1, 2.8, 2.6, and 2.8 percentage points higher than before the improvement. Comparative and ablation experiments have validated the proposed classification performance of this method, demonstrating its capability to facilitate an accurate and efficient automated auxiliary diagnosis of AD, offering a deep learning-based solution for it.

Research on Software Defect Prediction Model based on Deep Learning

As software systems grow in complexity and scale, detecting and predicting defects has become crucial for ensuring software quality and enhancing development efficiency. Traditional approaches to software defect prediction rely heavily on manual feature extraction and statistical models, which often struggle to handle intricate defect patterns and large-scale datasets. Recently, deep learning has demonstrated significant promise in software defect prediction, primarily due to its ability to automatically extract features and its strong pattern recognition capabilities. To enhance both the accuracy of defect predictions and the interpretability of the models, this paper proposes a deep learning-based prediction model. The focus is placed on the stages of data preprocessing, model selection, parameter tuning, and model assessment. The experimental outcomes indicate that the deep learning approach consistently outperforms conventional methods across various public datasets, achieving superior predictive accuracy and robustness. This study offers theoretical insights and practical evidence for further advancing the effectiveness of software defect prediction techniques.

Weighted Linear Discriminant Analysis: An Effective Feature Extraction Method for Multi-Class Imbalanced Datasets

In high-dimensional machine learning tasks, supervised feature extraction is essential for improving model performance, with Linear Discriminant Analysis (LDA) being a common approach. However, LDA tends to deliver suboptimal performance when dealing with class imbalance. To address this issue, we propose a novel feature extraction model, Weighted Linear Discriminant Analysis (WLDA), which integrates cost-sensitive techniques into the traditional LDA framework. By assigning weights inversely proportional to class sample sizes, WLDA achieves effective feature extraction under imbalanced sample conditions. We introduce an efficient solution algorithm for the proposed model and provide a thorough complexity analysis. Experimental results demonstrate the superior performance of WLDA in handling imbalanced datasets, confirming its potential as a robust tool for high-dimensional data scenarios. Overall, WLDA not only improves feature extraction for imbalanced datasets but also enhances classification accuracy across diverse applications.