A Semantic Segmentation Method for Remote Sensing Images Based on an Improved TransDeepLab Model DOI Creative Commons
Jinxin Wang, Manman Wang,

Kaiwei Cong

et al.

Land, Journal Year: 2024, Volume and Issue: 14(1), P. 22 - 22

Published: Dec. 26, 2024

Due to the various types of land cover and large spectral differences in remote sensing images, high-quality semantic segmentation these images still faces challenges such as fuzzy object boundary extraction difficulty identifying small targets. To address challenges, this study proposes a new improved model based on TransDeepLab method. The introduces GAM attention mechanism coding stage, incorporates multi-level linear up-sampling strategy decoding stage. These enhancements allow fully utilize information target details high-resolution thereby effectively improving accuracy objects. Using open-source LoveDA image datasets for validation experiment, results show that compared original model, model’s MIOU increased by 2.68%, aACC 3.41%, mACC 4.65%. Compared other mainstream models, also achieved superior performance.

Language: Английский

Visual fire detection using deep learning: A survey DOI
Guangtao Cheng, Xue Chen, Chenyi Wang

et al.

Neurocomputing, Journal Year: 2024, Volume and Issue: 596, P. 127975 - 127975

Published: June 1, 2024

Language: Английский

Citations

11
Fire and Smoke Detection Using Fine-Tuned YOLOv8 and YOLOv7 Deep Models DOI Creative Commons
Mohamed Chetoui, Moulay A. Akhloufi

Fire, Journal Year: 2024, Volume and Issue: 7(4), P. 135 - 135

Published: April 12, 2024

Viewed as a significant natural disaster, wildfires present serious threat to human communities, wildlife, and forest ecosystems. The frequency of wildfire occurrences has increased recently, with the impacts global warming interaction environment playing pivotal roles. Addressing this challenge necessitates ability firefighters promptly identify fires based on early signs smoke, allowing them intervene prevent further spread. In work, we adapted optimized recent deep learning object detection, namely YOLOv8 YOLOv7 models, for detection smoke fire. Our approach involved utilizing dataset comprising over 11,000 images fires. models successfully identified fire achieving mAP:50 92.6%, precision score 83.7%, recall 95.2%. results were compared YOLOv6 large model, Faster-RCNN, DEtection TRansformer. obtained scores confirm potential proposed wide application promotion in safety industry.

Language: Английский

Citations

9
YOLOGX: an improved forest fire detection algorithm based on YOLOv8 DOI Creative Commons

Caixiong Li,

Yue Du, Xing Zhang

et al.

Frontiers in Environmental Science, Journal Year: 2025, Volume and Issue: 12

Published: Jan. 7, 2025

To tackle issues, including environmental sensitivity, inadequate fire source recognition, and inefficient feature extraction in existing forest detection algorithms, we developed a high-precision algorithm, YOLOGX. YOLOGX integrates three pivotal technologies: First, the GD mechanism fuses extracts features from multi-scale information, significantly enhancing capability for targets of varying sizes. Second, SE-ResNeXt module is integrated into head, optimizing capability, reducing number parameters, improving accuracy efficiency. Finally, proposed Focal-SIoU loss function replaces original function, effectively directional errors by combining angle, distance, shape, IoU losses, thus model training process. was evaluated on D-Fire dataset, achieving [email protected] 80.92% speed 115 FPS, surpassing most classical algorithms specialized models. These enhancements establish as robust efficient solution detection, providing significant improvements reliability.

Language: Английский

Citations

0
ES-YOLOv8: a real-time defect detection algorithm in transmission line insulators DOI

Xiaoyang Song,

Qianlai Sun,

Jiayao Liu

et al.

Journal of Real-Time Image Processing, Journal Year: 2025, Volume and Issue: 22(2)

Published: Feb. 28, 2025

Language: Английский

Citations

0
Forest Fire Prediction Based on Time Series Networks and Remote Sensing Images DOI Open Access
Yue Cao,

Xuanyu Zhou,

Yanqi Yu

et al.

Forests, Journal Year: 2024, Volume and Issue: 15(7), P. 1221 - 1221

Published: July 14, 2024

Protecting forest resources and preventing fires are vital for social development public well-being. However, current research studies on fire warning systems often focus extensive geographic areas like states, counties, provinces. This approach lacks the precision detail needed predicting in smaller regions. To address this gap, we propose a Transformer-based time series forecasting model aimed at improving accuracy of predictions areas. Our study focuses Quanzhou County, Guilin City, Guangxi Province, China. We utilized data from 2021 to 2022, along with remote sensing images ArcGIS technology, identify various factors influencing region. established dataset containing twelve factors, each labeled occurrences. By integrating these Transformer model, generated danger level prediction maps County. model’s performance is compared other deep learning methods using metrics such as RMSE, results reveal that proposed achieves higher (ACC = 0.903, MAPE 0.259, MAE 0.053, RMSE 0.389). demonstrates effectively takes advantage spatial background information periodicity significantly enhancing predictive accuracy.

Language: Английский

Citations

3
YOLO-SIFD: YOLO with Sliced Inference and Fractal Dimension Analysis for Improved Fire and Smoke Detection DOI Open Access

Mr. Muhammad Ishtiaq,

Jong-Un Won

Computers, materials & continua/Computers, materials & continua (Print), Journal Year: 2025, Volume and Issue: 82(3), P. 5343 - 5361

Published: Jan. 1, 2025

Language: Английский

Citations

0
One-Year-Old Precocious Chinese Mitten Crab Identification Algorithm Based on Task Alignment DOI Creative Commons
Hao Gu,

Dongmei Gan,

Ming Chen

et al.

Animals, Journal Year: 2024, Volume and Issue: 14(14), P. 2128 - 2128

Published: July 21, 2024

The cultivation of the Chinese mitten crab (

Language: Английский

Citations

1
CL-YOLOv8: Crack Detection Algorithm for Fair-Faced Walls Based on Deep Learning DOI Creative Commons
Qinjun Li, Guoyu Zhang, Ping Yang

et al.

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(20), P. 9421 - 9421

Published: Oct. 16, 2024

Cracks pose a critical challenge in the preservation of historical buildings worldwide, particularly fair-faced walls, where timely and accurate detection is essential to prevent further degradation. Traditional image processing methods have proven inadequate for effectively detecting building cracks. Despite global advancements deep learning, crack under diverse environmental lighting conditions remains significant technical hurdle, as highlighted by recent international studies. To address this challenge, we propose an enhanced algorithm, CL-YOLOv8 (ConvNeXt V2-LSKA-YOLOv8). By integrating well-established ConvNeXt V2 model backbone network into YOLOv8, algorithm benefits from advanced feature extraction techniques, leading superior accuracy. This choice leverages V2’s recognized strengths, providing robust foundation improving overall performance. Additionally, introducing LSKA (Large Separable Kernel Attention) mechanism SPPF structure, receptive field enlarged correlations are strengthened, enhancing accuracy environments. study also contributes significantly expanding dataset wall detection, increasing its size sevenfold through data augmentation inclusion additional data. Our experimental results demonstrate that outperforms mainstream algorithms such Faster R-CNN, YOLOv5s, YOLOv7-tiny, SSD, various YOLOv8n/s/m/l/x models. achieves 85.3%, recall rate 83.2%, mean average precision (mAP) 83.7%. Compared YOLOv8n base model, shows improvements 0.9%, 2.3%, 3.9% accuracy, rate, mAP, respectively. These underscore effectiveness superiority positioning it valuable tool effort preserve architectural heritage.

Language: Английский

Citations

1
Fire Video Recognition Based on Channel Feature Enhancement DOI

健 丁

Artificial Intelligence and Robotics Research, Journal Year: 2024, Volume and Issue: 13(02), P. 185 - 193

Published: Jan. 1, 2024

Language: Английский

Citations

0
A Semantic Segmentation Method for Remote Sensing Images Based on an Improved TransDeepLab Model DOI Creative Commons
Jinxin Wang, Manman Wang,

Kaiwei Cong

et al.

Land, Journal Year: 2024, Volume and Issue: 14(1), P. 22 - 22

Published: Dec. 26, 2024

Due to the various types of land cover and large spectral differences in remote sensing images, high-quality semantic segmentation these images still faces challenges such as fuzzy object boundary extraction difficulty identifying small targets. To address challenges, this study proposes a new improved model based on TransDeepLab method. The introduces GAM attention mechanism coding stage, incorporates multi-level linear up-sampling strategy decoding stage. These enhancements allow fully utilize information target details high-resolution thereby effectively improving accuracy objects. Using open-source LoveDA image datasets for validation experiment, results show that compared original model, model’s MIOU increased by 2.68%, aACC 3.41%, mACC 4.65%. Compared other mainstream models, also achieved superior performance.

Language: Английский

Citations

0