Diagnosing gastrointestinal tract diseases is a critical task requiring accurate and efficient methodologies.While deep learning models have significantly advanced medical image analysis,challenges such as imbalanced ...Diagnosing gastrointestinal tract diseases is a critical task requiring accurate and efficient methodologies.While deep learning models have significantly advanced medical image analysis,challenges such as imbalanced datasets and redundant features persist.This study proposes a novel framework that customizes two deep learning models,NasNetMobile and ResNet50,by incorporating bottleneck architectures,named as NasNeck and ResNeck,to enhance feature extraction.The feature vectors are fused into a combined vector,which is further optimized using an improved Whale Optimization Algorithm to minimize redundancy and improve discriminative power.The optimized feature vector is then classified using artificial neural network classifiers,effectively addressing the limitations of traditional methods.Data augmentation techniques are employed to tackle class imbalance,improving model learning and generalization.The proposed framework was evaluated on two publicly available datasets:Hyper-Kvasir and Kvasir v2.The Hyper-Kvasir dataset,comprising 23 gastrointestinal disease classes,yielded an impressive 96.0%accuracy.On the Kvasir v2 dataset,which contains 8 distinct classes,the framework achieved a remarkable 98.9%accuracy,further demonstrating its robustness and superior classification performance across different gastrointestinal datasets.The results demonstrate the effectiveness of customizing deep models with bottleneck architectures,feature fusion,and optimization techniques in enhancing classification accuracy while reducing computational complexity.展开更多
Human action recognition(HAR)based on Artificial intelligence reasoning is the most important research area in computer vision.Big breakthroughs in this field have been observed in the last few years;additionally,the ...Human action recognition(HAR)based on Artificial intelligence reasoning is the most important research area in computer vision.Big breakthroughs in this field have been observed in the last few years;additionally,the interest in research in this field is evolving,such as understanding of actions and scenes,studying human joints,and human posture recognition.Many HAR techniques are introduced in the literature.Nonetheless,the challenge of redundant and irrelevant features reduces recognition accuracy.They also faced a few other challenges,such as differing perspectives,environmental conditions,and temporal variations,among others.In this work,a deep learning and improved whale optimization algorithm based framework is proposed for HAR.The proposed framework consists of a few core stages i.e.,frames initial preprocessing,fine-tuned pre-trained deep learning models through transfer learning(TL),features fusion using modified serial based approach,and improved whale optimization based best features selection for final classification.Two pre-trained deep learning models such as InceptionV3 and Resnet101 are fine-tuned and TL is employed to train on action recognition datasets.The fusion process increases the length of feature vectors;therefore,improved whale optimization algorithm is proposed and selects the best features.The best selected features are finally classified usingmachine learning(ML)classifiers.Four publicly accessible datasets such as Ut-interaction,Hollywood,Free Viewpoint Action Recognition usingMotion History Volumes(IXMAS),and centre of computer vision(UCF)Sports,are employed and achieved the testing accuracy of 100%,99.9%,99.1%,and 100%respectively.Comparison with state of the art techniques(SOTA),the proposed method showed the improved accuracy.展开更多
X-Ray knee imaging is widely used to detect knee osteoarthritis due to ease of availability and lesser cost.However,the manual categorization of knee joint disorders is time-consuming,requires an expert person,and is ...X-Ray knee imaging is widely used to detect knee osteoarthritis due to ease of availability and lesser cost.However,the manual categorization of knee joint disorders is time-consuming,requires an expert person,and is costly.This article proposes a new approach to classifying knee osteoarthritis using deep learning and a whale optimization algorithm.Two pre-trained deep learning models(Efficientnet-b0 and Densenet201)have been employed for the training and feature extraction.Deep transfer learning with fixed hyperparameter values has been employed to train both selected models on the knee X-Ray images.In the next step,fusion is performed using a canonical correlation approach and obtained a feature vector that has more information than the original feature vector.After that,an improved whale optimization algorithm is developed for dimensionality reduction.The selected features are finally passed to the machine learning algorithms such as Fine-Tuned support vector machine(SVM)and neural networks for classification purposes.The experiments of the proposed framework have been conducted on the publicly available dataset and obtained the maximum accuracy of 90.1%.Also,the system is explained using Explainable Artificial Intelligence(XAI)technique called occlusion,and results are compared with recent research.Based on the results compared with recent techniques,it is shown that the proposed method’s accuracy significantly improved.展开更多
Human Suspicious Activity Recognition(HSAR)is a critical and active research area in computer vision that relies on artificial intelligence reasoning.Significant advances have been made in this field recently due to i...Human Suspicious Activity Recognition(HSAR)is a critical and active research area in computer vision that relies on artificial intelligence reasoning.Significant advances have been made in this field recently due to important applications such as video surveillance.In video surveillance,humans are monitored through video cameras when doing suspicious activities such as kidnapping,fighting,snatching,and a few more.Although numerous techniques have been introduced in the literature for routine human actions(HAR),very few studies are available for HSAR.This study proposes a deep convolutional neural network(CNN)and optimal featuresbased framework for HSAR in video frames.The framework consists of various stages,including preprocessing video frames,fine-tuning deep models(Darknet 19 and Nasnet mobile)using transfer learning,serial-based feature fusion,feature selection via equilibrium feature optimizer,and neural network classifiers for classification.Fine-tuning two models using some hit and trial methods is the first challenge of this work that was later employed for feature extraction.Next,features are fused in a serial approach,and then an improved optimization method is proposed to select the best features.The proposed technique was evaluated on two action datasets,Hybrid-KTH01 and HybridKTH02,and achieved an accuracy of 99.8%and 99.7%,respectively.The proposed method exhibited higher precision compared to existing state-ofthe-art approaches.展开更多
基金supported by Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia through the Researchers Supporting Project PNURSP2025R333.
文摘Diagnosing gastrointestinal tract diseases is a critical task requiring accurate and efficient methodologies.While deep learning models have significantly advanced medical image analysis,challenges such as imbalanced datasets and redundant features persist.This study proposes a novel framework that customizes two deep learning models,NasNetMobile and ResNet50,by incorporating bottleneck architectures,named as NasNeck and ResNeck,to enhance feature extraction.The feature vectors are fused into a combined vector,which is further optimized using an improved Whale Optimization Algorithm to minimize redundancy and improve discriminative power.The optimized feature vector is then classified using artificial neural network classifiers,effectively addressing the limitations of traditional methods.Data augmentation techniques are employed to tackle class imbalance,improving model learning and generalization.The proposed framework was evaluated on two publicly available datasets:Hyper-Kvasir and Kvasir v2.The Hyper-Kvasir dataset,comprising 23 gastrointestinal disease classes,yielded an impressive 96.0%accuracy.On the Kvasir v2 dataset,which contains 8 distinct classes,the framework achieved a remarkable 98.9%accuracy,further demonstrating its robustness and superior classification performance across different gastrointestinal datasets.The results demonstrate the effectiveness of customizing deep models with bottleneck architectures,feature fusion,and optimization techniques in enhancing classification accuracy while reducing computational complexity.
基金This research work is supported in part by Chiang Mai University and HITEC University.
文摘Human action recognition(HAR)based on Artificial intelligence reasoning is the most important research area in computer vision.Big breakthroughs in this field have been observed in the last few years;additionally,the interest in research in this field is evolving,such as understanding of actions and scenes,studying human joints,and human posture recognition.Many HAR techniques are introduced in the literature.Nonetheless,the challenge of redundant and irrelevant features reduces recognition accuracy.They also faced a few other challenges,such as differing perspectives,environmental conditions,and temporal variations,among others.In this work,a deep learning and improved whale optimization algorithm based framework is proposed for HAR.The proposed framework consists of a few core stages i.e.,frames initial preprocessing,fine-tuned pre-trained deep learning models through transfer learning(TL),features fusion using modified serial based approach,and improved whale optimization based best features selection for final classification.Two pre-trained deep learning models such as InceptionV3 and Resnet101 are fine-tuned and TL is employed to train on action recognition datasets.The fusion process increases the length of feature vectors;therefore,improved whale optimization algorithm is proposed and selects the best features.The best selected features are finally classified usingmachine learning(ML)classifiers.Four publicly accessible datasets such as Ut-interaction,Hollywood,Free Viewpoint Action Recognition usingMotion History Volumes(IXMAS),and centre of computer vision(UCF)Sports,are employed and achieved the testing accuracy of 100%,99.9%,99.1%,and 100%respectively.Comparison with state of the art techniques(SOTA),the proposed method showed the improved accuracy.
基金supported by“Human Resources Program in Energy Technology”of the Korea Institute of Energy Technology Evaluation and Planning (KETEP),granted financial resources from the Ministry of Trade,Industry&Energy,Republic of Korea.(No.20204010600090).
文摘X-Ray knee imaging is widely used to detect knee osteoarthritis due to ease of availability and lesser cost.However,the manual categorization of knee joint disorders is time-consuming,requires an expert person,and is costly.This article proposes a new approach to classifying knee osteoarthritis using deep learning and a whale optimization algorithm.Two pre-trained deep learning models(Efficientnet-b0 and Densenet201)have been employed for the training and feature extraction.Deep transfer learning with fixed hyperparameter values has been employed to train both selected models on the knee X-Ray images.In the next step,fusion is performed using a canonical correlation approach and obtained a feature vector that has more information than the original feature vector.After that,an improved whale optimization algorithm is developed for dimensionality reduction.The selected features are finally passed to the machine learning algorithms such as Fine-Tuned support vector machine(SVM)and neural networks for classification purposes.The experiments of the proposed framework have been conducted on the publicly available dataset and obtained the maximum accuracy of 90.1%.Also,the system is explained using Explainable Artificial Intelligence(XAI)technique called occlusion,and results are compared with recent research.Based on the results compared with recent techniques,it is shown that the proposed method’s accuracy significantly improved.
基金supported by the“Human Resources Program in Energy Technology”of the Korea Institute of Energy Technology Evaluation and Planning (KETEP)and granted financial resources from the Ministry of Trade,Industry,and Energy,Republic of Korea (No.20204010600090).
文摘Human Suspicious Activity Recognition(HSAR)is a critical and active research area in computer vision that relies on artificial intelligence reasoning.Significant advances have been made in this field recently due to important applications such as video surveillance.In video surveillance,humans are monitored through video cameras when doing suspicious activities such as kidnapping,fighting,snatching,and a few more.Although numerous techniques have been introduced in the literature for routine human actions(HAR),very few studies are available for HSAR.This study proposes a deep convolutional neural network(CNN)and optimal featuresbased framework for HSAR in video frames.The framework consists of various stages,including preprocessing video frames,fine-tuning deep models(Darknet 19 and Nasnet mobile)using transfer learning,serial-based feature fusion,feature selection via equilibrium feature optimizer,and neural network classifiers for classification.Fine-tuning two models using some hit and trial methods is the first challenge of this work that was later employed for feature extraction.Next,features are fused in a serial approach,and then an improved optimization method is proposed to select the best features.The proposed technique was evaluated on two action datasets,Hybrid-KTH01 and HybridKTH02,and achieved an accuracy of 99.8%and 99.7%,respectively.The proposed method exhibited higher precision compared to existing state-ofthe-art approaches.
