The COVID-19 pandemic has been associated with several manifestations that affect an individual’s visual health. To address these concerns, we propose a classification model to predict the impact of COVID-19 on visio...
The COVID-19 pandemic has been associated with several manifestations that affect an individual’s visual health. To address these concerns, we propose a classification model to predict the impact of COVID-19 on vision. The model is built on a dataset corresponding to visual symptoms and other factors. The proposed classification model will classify people into three categories: those with no impact on vision, those with mild-to-moderate impact, and those with severe impact. Some machine learning algorithms, such as logistic regression, decision trees, random forests, etc., were applied by us to build the model to identify the optimal algorithm for the task. The results obtained from the model show that it has high accuracy, precision, and recall. The model can predict the severity of visual symptoms in people due to COVID-19 with an accuracy of over 84%. The study findings are entirely focused on the E-learning or online learning caused due to COVID-19 pandemic and their impact on vision of people from different age groups.
In recent years, the need for automatic detection of Not Safe for Work (NSFW) content on social media platforms has increased dramatically. In this study, we contrasted the presentation of five optimizers, namely ADAM...
In recent years, the need for automatic detection of Not Safe for Work (NSFW) content on social media platforms has increased dramatically. In this study, we contrasted the presentation of five optimizers, namely ADAM, ADAMAX, ADAMW, SGD, and ADAGRAD, on the EfficientNet-V2M and V2L models for NSFW content detection. We used a dataset consisting of NSFW images for training and testing the models. The results show that the ADAM optimizer performed better than the other optimizers with an accuracy of 98.80% for training and that for testing is 95.60% for the EfficientNet-V2L model. However, all the optimizers performed reasonably well with same parameters. This study provides valuable insights into the selection of optimizers for NSFW content detection using deep learning models. The dataset used in the research consists of a large number of images with explicit and non-explicit content. The results were evaluated based on accuracy, precision, recall, F1-score, loss, and area under the curve (AUC). The findings indicate that ADAM, ADAMAX, and ADAMW outperform the other optimizers in terms of all evaluation metrics. With the easy availability of explicit content online, it has become a concern for society, especially for children who are easily influenced by such content on various platforms. Exposure to unfiltered and inappropriate content can have a negative impact on young minds. To safeguard against such content, the authors of this paper review and analyze different approaches for detecting and filtering pornographic and NSFW content. The objective of this paper is to provide a filtered and safe content environment for the community, especially for children and teenagers who are the most vulnerable.
Speech is the primary form of communication, and being able to identify the language used in an audio sample is crucial in numerous fields. Language identification serves a purpose in speech recognition, language tran...
Speech is the primary form of communication, and being able to identify the language used in an audio sample is crucial in numerous fields. Language identification serves a purpose in speech recognition, language translation, voice assistants, and many other domains. Traditional language identification systems rely on statistical or acoustic models, which frequently require extensive domain-specific knowledge and have limitations in accuracy and resilience. By employing a hybrid Recurrent neural network and Convolutional Neural Network, this paper aims to develop a novel method of language identification. The suggested approach entails spectrogram preprocessing of audio samples, Mel-Frequency cepstrum coefficient extraction, and Convolutional Neural Network and Recurrent neural network architecture model training on a sizable dataset. Recurrent neural networks are good at capturing temporal dependencies in data, while Convolutional Neural Networks are better at capturing spatial patterns. This methodology successfully identifies the languages with an overall accuracy of 93%, proving the efficacy of the proposed model.
We present a Multi Lingual Sync model for generating lip-synced videos in multiple languages. The model consists of Lingua Speak for translation and Wav2Lip for lip synchronization. The workflow involves extracting au...
We present a Multi Lingual Sync model for generating lip-synced videos in multiple languages. The model consists of Lingua Speak for translation and Wav2Lip for lip synchronization. The workflow involves extracting audio from a video, translating it using Lingua Speak, converting the translated text to speech, and then using Wav2Lip to generate synchronized lip movements. Wav2Lip utilizes a Generator with an identity encoder, speech encoder, and face decoder, along with a pre-trained lip-sync discriminator called SyncNet. The proposed model enables the creation of accurate and efficient multilingual videos with synchronized speech and video.
Education in recent years has slowly transitioned to an online model, allowing massive access to online courses virtually from anywhere. The adoption of such educational models was boosted by the global pandemic in 20...
Education in recent years has slowly transitioned to an online model, allowing massive access to online courses virtually from anywhere. The adoption of such educational models was boosted by the global pandemic in 2020, with universities and other degree programs quickly transitioning to such schemes. Although such a model is apt for lecture-based courses, hands-on training remains a puzzle on how it can transition to remote learning. In this work, we describe and evaluate our scheme for integrating testbed resources in online-taught networking-related courses in University of Thessaly, Greece. The framework is based on Kubernetes and is able to deliver hands-on labs related to networking as micro-services over the testbed architecture with minimal overhead on the lab setup from the instructor. The proposed approach has been applied in the networking-related courses of the curriculum during the 2020-2021 and 2021-2022 academic years, educating more than 800 students on computer networking concepts in practice. The paper describes the framework and a benchmarking evaluation, which proves the capacity of the framework to serve up to 5 times higher numbers of students, compared to prior methodologies and practices, without any infrastructure upgrades.
Through the use of the Fundamental Lemma for linear systems, a direct data-driven state-feedback control synthesis method is presented for a rather general class of nonlinear (NL) systems. The core idea is to develop ...
详细信息
Speech recognition, a disruptive technology, has revolutionized human-machine interaction. While numerous Automatic Speech Recognition (ASR) models are publicly available via HuggingFace, the majority cater to English...
Speech recognition, a disruptive technology, has revolutionized human-machine interaction. While numerous Automatic Speech Recognition (ASR) models are publicly available via HuggingFace, the majority cater to English language. For Urdu, however, models are scarce or closed-source, with open-sourced ones often lack the robustness. Our research addresses this scarcity, focusing on the challenges posed by dialects, slangs, and accents. In this work, we introduce an innovative ASR model leveraging the pretrained XLS-R model based on Wav2Vec2.0 architecture, trained on CommonVoice corpus V11. Our approach outperforms other deep learning-based techniques both qualitatively and quantitatively, offering promising results for an under-resourced language.
Maintaining strong network security in the quickly developing Internet of Things (IoT) space is crucial because IoT traffic is so varied and complicated. By focusing on the TON IoT dataset, this study introduces a nov...
详细信息
ISBN:
(数字)9798331519094
ISBN:
(纸本)9798331519100
Maintaining strong network security in the quickly developing Internet of Things (IoT) space is crucial because IoT traffic is so varied and complicated. By focusing on the TON IoT dataset, this study introduces a novel method for intrusion detection in IoT networks. Bidirectional Gated Recurrent Units (Bi-GRU) were utilized in a deep learning architecture to capture temporal dependencies in the data, while Convolutional Neural Networks (CNN) were utilized for effective feature extraction. By including a Multi-Head Attention layer, the model was able to select and pay attention to important characteristics, which enhanced its ability to focus on important patterns. A hybrid feature selection approach incorporating MI, Lasso, and RFECV was employed to generate a refined feature set, maintaining critical information for improved model performance. With a 99.73% binary classification accuracy and a 99.68% multi-class classification accuracy, our work outperformed the state-of-the-art techniques. These findings highlight how well CNN, Bi-GRU, and Attention mechanisms work together to detect intricate intrusion patterns in IoT environments, which advances the security measures required to shield IoT networks from dynamic cyber attacks.
Lung cancer is a major contributor to global mortality rates and identification is critical to improve patient outcomes. In recent years, machine learning algorithms have demonstrated promising results in identifying ...
Lung cancer is a major contributor to global mortality rates and identification is critical to improve patient outcomes. In recent years, machine learning algorithms have demonstrated promising results in identifying lung nodules from medical images. The most compelling area of research for scientists is the early detection of lung cancer. This study is a method for lung nodule detection using CT images. The study incorporates a hybrid model that combines multiple machine learning algorithms including CNN, SVM, DTC, ANN, and KNN to improve the accuracy of nodule detection. The hybrid model demonstrated high accuracy in identifying various types of lung nodules, including Adenocell carcinoma, squamous cell carcinoma, and large cell carcinoma. Specifically, the model achieved an accuracy rate of over 90% in detecting and differentiating normal lung tissue and Adenocell carcinomas. Accuracy graphs and priority setting were utilized to assess the model's capability in accurately predicting the presence of lung cancer. Additionally, the efficiency of the hybrid model was compared with other machine learning algorithms, including SVM, Random Forest, and Decision Trees. A large dataset of CT scans was collected for training and evaluation purposes. The results demonstrated the advantages of the suggested hybrid model in terms of accuracy and efficiency. This study highlights the importance of early lung nodule identification using CT scans and demonstrates the effectiveness of the hybrid model in accurately identifying different types of lung nodules.
With increasing camera enabled devices huge video data is generated every second. With unlimited storage offered by cloud, most of this video data is moved to cloud storage directly or indirectly. But manually queryin...
详细信息
暂无评论