An Attention-Guided Deep-Learning-Based Network with Bayesian Optimization for Forest Fire Classification and Localization

Islam, Al Mohimanul; Masud, Fatiha Binta; Ahmed, Md. Rayhan; Jafar, Anam Ibn; Ullah, Jeath Rahmat; Islam, Salekul; Shatabda, Swakkhar; Islam, A. K. M. Muzahidul

doi:10.3390/f14102080

Cited by 4 publications

(3 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally, we have conducted an error analysis based on feature visualization using gradient-weighted class activation mapping (Grad-CAM) [39], following methodologies outlined in References [15,40]. This qualitative analysis highlights the improved ability of the proposed approach in accurately identifying distinguishing image features.…”

Section: Metricsmentioning

confidence: 99%

Dual-Dataset Deep Learning for Improved Forest Fire Detection: A Novel Hierarchical Domain-Adaptive Learning Approach

El-Madafri,

Peña,

Olmedo-Torre

2024

Mathematics

View full text Add to dashboard Cite

This study introduces a novel hierarchical domain-adaptive learning framework designed to enhance wildfire detection capabilities, addressing the limitations inherent in traditional convolutional neural networks across varied forest environments. The framework innovatively employs a dual-dataset approach, integrating both non-forest and forest-specific datasets to train a model adept at handling diverse wildfire scenarios. The methodology leverages a novel framework that combines shared layers for broad feature extraction with specialized layers for forest-specific details, demonstrating versatility across base models. Initially demonstrated with EfficientNetB0, this adaptable approach could be applicable with various advanced architectures, enhancing wildfire detection. The research’s comparative analysis, benchmarking against conventional methodologies, showcases the proposed approach’s enhanced performance. It particularly excels in accuracy, precision, F1-score, specificity, MCC, and AUC-ROC. This research significantly reduces false positives in wildfire detection through a novel blend of multi-task learning, dual-dataset training, and hierarchical domain adaptation. Our approach advances deep learning in data-limited, complex environments, offering a critical tool for ecological conservation and community protection against wildfires.

show abstract

Section: Metricsmentioning

confidence: 99%

Dual-Dataset Deep Learning for Improved Forest Fire Detection: A Novel Hierarchical Domain-Adaptive Learning Approach

El-Madafri,

Peña,

Olmedo-Torre

2024

Mathematics

View full text Add to dashboard Cite

show abstract

“…In [5], the authors propose a multi-stream hybrid deep learning model with a dualstream attention mechanism for classifying wildfires from aerial and territorial images. The model is based on a pre-trained EfficientNetB7 and a customized Attention Connected Network with Bayesian optimization.…”

Section: Related Workmentioning

confidence: 99%

Intelligent Methods for Forest Fire Detection Using Unmanned Aerial Vehicles

Abramov,

Emelyanova,

Fralenko

et al. 2024

Fire

View full text Add to dashboard Cite

This research addresses the problem of early detection of smoke and open fire on the observed territory by unmanned aerial vehicles. We solve the tasks of improving the quality of incoming video data by removing motion blur and stabilizing the video stream; detecting the horizon line in the frame; and identifying fires using semantic segmentation with Euclidean–Mahalanobis distance and the modified convolutional neural network YOLO. The proposed horizon line detection algorithm allows for cutting off unnecessary information such as cloud-covered areas in the frame by calculating local contrast, which is equivalent to the pixel informativeness indicator of the image. Proposed preprocessing methods give a delay of no more than 0.03 s due to the use of a pipeline method for data processing. Experimental results show that the horizon clipping algorithm improves fire and smoke detection accuracy by approximately 11%. The best results with the neural network were achieved with YOLO 5m, which yielded an F1 score of 76.75% combined with a processing speed of 45 frames per second. The obtained results differ from existing analogs by utilizing a comprehensive approach to early fire detection, which includes image enhancement and alternative real-time video processing methods.

show abstract

“…In the meantime, the hyperparameter selection of EfficientNetB7 is performed by applying the EO model. The EO algorithm is based on the formula of mass balance in a control volume from the physical principles, trying to find out the equilibrium state of the system [21]. Initialization, equilibrium pooling, and concentration update are the three stages of EO.…”

Section: Parameter Tuning Using the Eo Modelmentioning

confidence: 99%

Bioinspired Image Processing Enabled Facial Emotion Recognition Using Equilibrium Optimizer With a Hybrid Deep Learning Model

Alzahrani

2024

IEEE Access

View full text Add to dashboard Cite

Owing to the unpredictable nature of human facial expression, Facial emotion recognition (FER) from facial images becomes a tedious process. FER is a field within artificial intelligence (AI) and computer vision (CV) that concentrates on developing algorithms and technologies to interpret and analyze emotional expressions shown on human faces. The major intention of this area of research is to enable computers and machines to automatically classify and detect emotions, including sadness, happiness, anger, etc., based on facial expressions and cues. Deep learning (DL) systems namely recurrent neural network (RNN) and convolutional neural network (CNN) can be used for the task of automatically classifying and detecting emotions from human facial expressions. Furthermore, Bioinspired Image Processing Enabled FER combines principles from biological systems with image processing techniques to better the robustness and accuracy of FER. This technique is inspired by natural processes to enhance the interpretation and understanding of human emotion conveyed by facial expressions. This study designs a new Bioinspired Image Processing Enabled Facial Emotion Recognition using an Equilibrium Optimizer with Hybrid Deep Learning (BIPFER-EOHDL) model. The primary objectives of the BIPFER-EOHDL technique lie in the effectual and automated identification of facial expressions using a hyperparameter-tuned DL algorithm. In the presented BIPFER-EOHDL technique, the median filtering (MF) approach can be applied for image preprocessing. Besides, the BIPFER-EOHDL method applies the EfficientNetB7 model for the process of feature extraction. Meanwhile, the hyperparameter selection of the EfficientNetB7 model takes place by the use of the EO algorithm. Finally, the multi-head attention bi-directional long short-term memory (MA-BLSTM) model is exploited for the recognition and classification of facial emotions. A wide-ranging simulation analysis was performed to validate the higher FER outcomes of the BIPFER-EOHDL methodology. The simulation results stated that the BIPFER-EOHDL technique accomplishes better FER results than other recent approaches.

show abstract

An Attention-Guided Deep-Learning-Based Network with Bayesian Optimization for Forest Fire Classification and Localization

Cited by 4 publications

References 31 publications

Dual-Dataset Deep Learning for Improved Forest Fire Detection: A Novel Hierarchical Domain-Adaptive Learning Approach

Dual-Dataset Deep Learning for Improved Forest Fire Detection: A Novel Hierarchical Domain-Adaptive Learning Approach

Intelligent Methods for Forest Fire Detection Using Unmanned Aerial Vehicles

Bioinspired Image Processing Enabled Facial Emotion Recognition Using Equilibrium Optimizer With a Hybrid Deep Learning Model

Contact Info

Product

Resources

About