Human action recognition systems use data collected from a wide range of sensors to accurately identify and interpret human actions. One of the most challenging issues for computer vision is the automatic and precise identification of human activities. A significant increase in feature learning-based representations for action recognition has emerged in recent years, due to the widespread use of deep learning-based features. This study presents an in-depth analysis of human activity recognition that investigates recent developments in computer vision. Augmented reality, human–computer interaction, cybersecurity, home monitoring, and surveillance cameras are all examples of computer vision applications that often go in conjunction with human action detection. We give a taxonomy-based, rigorous study of human activity recognition techniques, discussing the best ways to acquire human action features, derived using RGB and depth data, as well as the latest research on deep learning and hand-crafted techniques. We also explain a generic architecture to recognize human actions in the real world and its current prominent research topic. At long last, we are able to offer some study analysis concepts and proposals for academics. In-depth researchers of human action recognition will find this review an effective tool.

Graph neural networks lose a lot of their computing power when more network layers are added. As a result, the majority of existing graph neural networks have a shallow depth of learning. Over-smoothing and information loss are two of the key issues that restrict graph neural networks from going deeper. As network depth goes up, the embeddings of all the nodes eventually converge on the same value, which separates output representations from input vectors and causes over-smoothing. Moreover, layers of graph pooling are required in a deep learning model to retrieve specified features for prediction, which results in some degree of information loss. In this research, we present a new and multi-scale approach for overcoming these constraints by using concepts from graph theory, namely learnable edge sampling and line graphs. An edge-sampling mechanism that selects a particular number of edges through a learning parameter before training reduces oversmoothing, and the issue of information loss is alleviated using a line graph technique that converts the original graph into a similar line graph. Our method of edge sampling preserves the core spectral features of the graph without affecting its fundamental structure. Our suggested technique outperforms state-of-the-art models on publicly available datasets of diverse applications while having minimal constraints and great training skills.

Flight delays continue to pose a substantial concern in the aviation sector, impacting both operational efficiency and passenger satisfaction. Existing systems, while attempting to predict delays, often lack accurate predictive capabilities due to poor modeling setups, insufficient feature engineering, and inadequate feature selection processes, leading to suboptimal predictions and ineffective decision-making. Precisely forecasting flight arrival delays is essential for improving airline scheduling and resource allocation. The aim of our research is to create a superior prediction model that surpasses current modeling approaches. This study aims to forecast airline arrival delays by examining data from five prominent U.S. states in 2023—California (CA), Texas (TX), Florida (FL), New York (NY), and Georgia (GA). Our proposed modeling approach involves feature engineering to identify significant variables, followed by a novel feature selection algorithm (CFS) designed to retain only the most relevant features. Delay forecasts were generated using our proposed Deep Feed Forward Regression Network (DFFRN), a five-layer deep learning approach designed to enhance predictive accuracy by incorporating extensively selected features. The findings indicate that the DFFRN model substantially outperformed conventional models documented in the literature. The DFFRN had the highest R² score (99.916%), indicating exceptional predictive efficacy, highlighting the efficacy of the DFFRN model for predicting flight delays and establishing it as a significant asset for improving decision-making and minimizing operational delays in the aviation sector.

Images, text, and other forms of data collected through the device or sensors that we carry with us, such as a smart phone, camera, and so on, are referred to as companion data. There is considerable interest in automatic assignment of keywords to images in order to index, retrieve and analyze huge amounts of image data. In the past decade, many image annotation methods have been developed, many of which perform well on benchmark datasets. However, very little attention has been given to annotating the companion image data in real time using deep neural network. A new baseline technique for companion image data annotation using deep neural networks is introduced in this study that approaches annotation as a retrieval task. As an object detector, the Single Shot Multibox Detector is utilized, while the MobileNetV2 architecture serves as the classifier. This portable and lightweight architecture may also be utilized in integrated systems (such as smart phones and Raspberry Pis) to conduct real-time image annotation. We achieved 82.37% precision and 74.03% F1 score using the method described in this paper. For image annotation, our baseline model beats the existing best practices. We anticipate that adopting a baseline will provide a solid foundation for comparing and comprehending future annotation methods.

Learning on graph-structured data is an area where graph neural networks (GNN) have gained widespread use. In several tasks, such as node classification and graph classification, they outperformed traditional heuristic techniques. When it comes to link prediction, where the edge features, particularly multi-dimensional edge data, are critical, GNNs generally perform poorly compared to simple heuristic approaches. In this research, we provide a novel method for graph neural networks family which can better explore edge characteristics. These features may include both directed and undirected edges, as well as edges with many dimensions. The suggested framework has the potential to unify existing models of graph neural networks like GCN and GAT. We build a new method for edge perturbation for every GNN layer which can process edge features with more than one dimension. We test our proposed model for graph link prediction on a wide range of publicly available graph datasets. Our proposed method surpass the existing stateof-the-art approaches, employed based on GCNs and GAT, demonstrating the significance of leveraging edge properties for graph neural networks.