Unraveling Overlying Rock Fracturing Evolvement for Mining Water Inflow Channel Prediction: A Spatiotemporal Analysis Using ConvLSTM Image Reconstruction

IEEE Transactions on Geoscience and Remote Sensing, Journal Year: 2024, Volume and Issue: 62, P. 1 - 17

Published: Jan. 1, 2024

Language: Английский

---

Experiments on image data augmentation techniques for geological rock type classification with convolutional neural networks

Journal of Rock Mechanics and Geotechnical Engineering, Journal Year: 2024, Volume and Issue: unknown

Published: May 1, 2024

The integration of image analysis through deep learning (DL) into rock classification represents a significant leap forward in geological research. While traditional methods remain invaluable for their expertise and historical context, DL offers powerful complement by enhancing the speed, objectivity, precision process. This research explores significance data augmentation techniques optimizing performance convolutional neural networks (CNNs) analysis, particularly igneous, metamorphic, sedimentary types from thin section (RTS) images. study primarily focuses on classic evaluates impact model accuracy precision. Results demonstrate that like Equalize significantly enhance model's capabilities, achieving an F1-Score 0.9869 igneous rocks, 0.9884 metamorphic 0.9929 representing improvements compared to baseline original results. Moreover, weighted average across all classes is 0.9886, indicating enhancement. Conversely, Distort lead decreased F1-Score, with 0.949 0.954 0.9416 exacerbating baseline. underscores practicality advocates adoption this domain automation improved findings can benefit various fields, including remote sensing, mineral exploration, environmental monitoring, both scientific industrial applications.

Language: Английский

---

Efficient surface crack segmentation for industrial and civil applications based on an enhanced YOLOv8 model

Journal Of Big Data, Journal Year: 2025, Volume and Issue: 12(1)

Published: Jan. 28, 2025

Abstract Crack segmentation is essential for preventive maintenance in various civil and industrial applications. It makes it possible to identify divide structural cracks or defects. Complicated sceneries, such as with an irregular form, complicated image environments, constraints obtaining global contextual information, affect the performance of crack segmentation. This research proposes Enhanced-YOLOv8 called YOLOv8-MHSA-TA reduce effects these factors offer quasi-real-time concurrent identification different types. The suggested network uses triplet attention (TA) multi-head self-attention (MHSA) mechanisms, enhance YOLOv8’s performance. To evaluate proposed approach test its generalization ability, nine public datasets comprising images structures were collected, including CracK500, Crack3238, Forest Dataset, Deepcrack, Rissbilder, Volker, Sylvie, Magnetic Tile, Pipeline Gamma Radiography Images. contain sizes, shapes, sorts, lighting situations, orientations. Applying enhanced YOLOv8 model’s capabilities, are detected segmented successfully examined images. results demonstrate that, Crack500 tile datasets, Mean Average Precision (mAP50) 10.1 26.4% higher than that original models. model was compared YOLOv8-MHSA, YOLOv8-TA, models, well other published networks. outcomes our outperforms previously work enhances method prior when considering diverse dataset.

Language: Английский

---

Automatic Detection and Classification of Dead Nematode-Infested Pine Wood in Stages Based on YOLO v4 and GoogLeNet

Forests, Journal Year: 2023, Volume and Issue: 14(3), P. 601 - 601

Published: March 17, 2023

Pine wood nematode disease has harmed forests in several countries, and can be reduced by locating clearing infested pine trees from forests. The target detection model of deep learning was utilized to monitor a nematode-infested wood. detecting effect good, but limited low-resolution photos with poor accuracy speed. Our work presents staged classification approach for dead based using You Only Look Once version 4 (YOLO v4) Google Inception 1 Net (GoogLeNet), employing high-resolution images acquired helicopter. Experiments showed that the method only YOLO v4 were comparable when amount data sufficient, former higher than latter. retained fast training speed one-stage model, further improving volume, more flexible achieving accurate classification, meeting needs forest areas epidemic prevention control.

Language: Английский

---

A new model based on improved VGG16 for corn weed identification

Frontiers in Plant Science, Journal Year: 2023, Volume and Issue: 14

Published: July 7, 2023

Weeds remain one of the most important factors affecting yield and quality corn in modern agricultural production. To use deep convolutional neural networks to accurately, efficiently, losslessly identify weeds fields, a new weed identification model, SE-VGG16, is proposed. The SE-VGG16 model uses VGG16 as basis adds SE attention mechanism realize that network automatically focuses on useful parts allocates limited information processing resources parts. Then 3 × kernels first block are reduced 1 kernels, ReLU activation function replaced by Leaky perform feature extraction while dimensionality reduction. Finally, it global average pooling layer for fully connected VGG16, output performed softmax. experimental results verify classifies superiorly other classical advanced multiscale models with an accuracy 99.67%, which more than 97.75% original model. Based three evaluation indices precision rate, recall F1, was concluded has good robustness, high stability, recognition can be used accurately provide effective solution control fields practical applications.

Language: Английский

---

Coordinate Attention Guided Dual-Teacher Adaptive Knowledge Distillation for image classification

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: 250, P. 123892 - 123892

Published: April 6, 2024

Language: Английский

---

Optimized Crop Disease Identification in Bangladesh: A Deep Learning and SVM Hybrid Model for Rice, Potato, and Corn

Journal of Imaging, Journal Year: 2024, Volume and Issue: 10(8), P. 183 - 183

Published: July 30, 2024

Agriculture plays a vital role in Bangladesh’s economy. It is essential to ensure the proper growth and health of crops for development agricultural sector. In context Bangladesh, crop diseases pose significant threat output and, consequently, food security. This necessitates timely precise identification such sustainability production. study focuses on building hybrid deep learning model three specific affecting major crops: late blight potatoes, brown spot rice, common rust corn. The proposed leverages EfficientNetB0′s feature extraction capabilities, known achieving rapid high rates, coupled with classification proficiency SVMs, well-established machine algorithm. unified approach streamlines data processing extraction, potentially improving generalizability across diverse diseases. also aims address challenges computational efficiency accuracy that are often encountered precision agriculture applications. achieved 97.29% accuracy. A comparative analysis other models, CNN, VGG16, ResNet50, Xception, Mobilenet V2, Autoencoders, Inception v3, EfficientNetB0 each an 86.57%, 83.29%, 68.79%, 94.07%, 90.71%, 87.90%, 94.14%, 96.14% respectively, demonstrated superior performance our model.

Language: Английский

---

Adaptive loss optimization for enhanced learning performance: application to image-based rock classification

Neural Computing and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 9, 2025

Language: Английский

---

A novel twin vision transformer framework for crop disease classification with deformable attention

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 105, P. 107551 - 107551

Published: Feb. 7, 2025

Language: Английский

---

Automatic smart brain tumor classification and prediction system using deep learning

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: April 28, 2025

A brain tumor is a serious medical condition characterized by the abnormal growth of cells within brain. It can cause range symptoms, including headaches, seizures, cognitive impairment, and changes in behavior. Brain tumors pose significant health concern, imposing substantial burden on patients. Timely diagnosis crucial for effective treatment patient health. be either benign or malignant, their symptoms often overlap with those other neurological conditions, leading to delays diagnosis. Early detection allow timely intervention, potentially preventing from reaching an advanced stage. This reduces risk complications increases rate recovery. also selection most suitable treatment. In recent years, Smart IoT devices deep learning techniques have brought remarkable success various imaging applications. study proposes smart monitoring system early detection, classification, prediction tumors. The proposed research employs custom CNN model two pre-trained models, specifically Inception-v4 EfficientNet-B4, classification cases into ten categories: Meningioma, Pituitary, No tumor, Astrocytoma, Ependymoma, Glioblastoma, Oligodendroglioma, Medulloblastoma, Germinoma, Schwannoma. designed focus computational efficiency adaptability address unique challenges classification. Its new makes it key component detection. Extensive experimentation conducted diverse set MRI datasets evaluate performance developed model. model's precision, sensitivity, accuracy, f1-score, error rate, specificity, Y-index, balanced geometric mean, ROC are considered as metrics. average accuracy CNN, Inception-v4, EfficientNet-B4 97.58%, 99.56%, 99.76%, respectively. results demonstrate excellent previous approaches. Furthermore, trained models maintain accurate after deployment. method predicts 96.5% 99.3% 99.7% test dataset 1000 images.

Language: Английский

---