Human action recognition (HAR), deeply rooted in computer vision, video surveillance, automated observation, and human computer interaction (HCI), enables precise identification of human actions. Numerous research groups have dedicated their efforts to various applications and problem domains in HAR systems. They trained classification models using diverse datasets, enhanced hardware capabilities, and employed different metrics to assess performance. Although several surveys and review articles have been published periodically to highlight research advancements in HAR, there is currently no comprehensive and up-to-date study that encompasses architecture, application areas, techniques/algorithms, and evaluation methods as well as challenges and issues. To bridge this gap in the literature, this article presents a comprehensive analysis of the current state of HAR systems by thoroughly examining a meticulously chosen collection of 136 publications published within the past two decades. These findings have significant implications for researchers engaged in different aspects of HAR systems.

Missing data is a significant issue in many real-world datasets, yet there are no robust methods for dealing with it appropriately. In this paper, we propose a robust approach to dealing with missing data in classification problems: Multiple Imputation Ensembles (MIE). Our method integrates two approaches: multiple imputation and ensemble methods and compares two types of ensembles: bagging and stacking. We also propose a robust experimental set-up using 20 benchmark datasets from the UCI machine learning repository. For each dataset, we introduce increasing amounts of data Missing Completely at Random. Firstly, we use a number of single/multiple imputation methods to recover the missing values and then ensemble a number of different classifiers built on the imputed data. We assess the quality of the imputation by using dissimilarity measures. We also evaluate the MIE performance by comparing classification accuracy on the complete and imputed data. Furthermore, we use the accuracy of simple imputation as a benchmark for comparison. We find that our proposed approach combining multiple imputation with ensemble techniques outperform others, particularly as missing data increases.

Human Action Recognition (HAR) is a vital area of computer vision with diverse applications in security, healthcare, and human-computer interaction. Addressing the challenges of HAR, particularly in dynamic and complex environments, is essential to advancing this field. The Human Actions in Diverse Environments (HADE) framework introduced in this paper represents a significant advancement in improving the capabilities of Convolutional Neural Networks (CNNs) for effective HAR. The strength of the HADE framework is its carefully curated dataset, primarily derived from smartphone camera footage. This dataset encompasses a wide range of human actions captured in various settings, providing a robust foundation for training our novel HAR models HADE I and HADE II. These models have been specifically designed and optimized for parallel processing on GPUs, showing significant improvements in the efficiency of both training and inference processes. Through a comprehensive evaluation, the HADE framework demonstrated a remarkable improvement in HAR accuracy, achieving 83.57% on our custom dataset. This marks a considerable enhancement over the existing methodologies and underscores the efficacy of the HADE approach in accurately interpreting complex human actions. Particularly noteworthy is the framework’s potential applicability in healthcare, in the domain of neurological patient care, where it can aid in early detection and facilitate personalized treatment plans. Future research will focus on expanding the range of actions covered by HAR and exploring avenues for real-time processing. The introduction of the HADE framework not only makes a substantial contribution to the field of computer vision but also paves the way for its practical application across various sectors.

In recent years, disease attacks have posed continuous threats to agriculture and caused substantial losses in the economy. Thus, early detection and classification could minimize the spread of disease and help to improve yield. Meanwhile, deep learning has emerged as the significant approach to detecting and classifying images. The classification performed using the deep learning approach mainly relies on large datasets to prevent overfitting problems. The Automatic Segmentation and Hyper Parameter Optimization Artificial Rabbits Algorithm (AS-HPOARA) is developed to overcome the above-stated issues. It aims to improve plant leaf disease classification. The Plant Village dataset is used to assess the proposed AS-HPOARA approach. Z-score normalization is performed to normalize the images using the dataset’s mean and standard deviation. Three augmentation techniques are used in this work to balance the training images: rotation, scaling, and translation. Before classification, image augmentation reduces overfitting problems and improves the classification accuracy. Modified UNet employs a more significant number of fully connected layers to better represent deeply buried characteristics; it is considered for segmentation. To convert the images from one domain to another in a paired manner, the classification is performed by HPO-based ARA, where the training data get increased and the statistical bias is eliminated to improve the classification accuracy. The model complexity is minimized by tuning the hyperparameters that reduce the overfitting issue. Accuracy, precision, recall, and F1 score are utilized to analyze AS-HPOARA’s performance. Compared to the existing CGAN-DenseNet121 and RAHC_GAN, the reported results show that the accuracy of AS-HPOARA for ten classes is high at 99.7%.

The remote sensing (RS) image classification task has been studied widely in the RS and geoscience community. The important applications of RS are landslides, earthquakes, land use, and land cover classification. Landslides and earthquakes are some of the most dangerous natural disasters that frequently occur. High-resolution remote sensing (HRRS) images can be useful for accurately classifying landslide and earthquake regions. The deep learning technique has improved performance compared to traditional methods; however, these techniques are reliable on large-scale datasets. In this work, we proposed a novel architecture based on super-resolution and fused bottleneck self-attention called (FMANet) convolutional neural network. A new custom deep super-resolution network is designed as the first step to improve the quality of RS images. In the next step, a new fused bottleneck self-attention architecture is proposed that learns the features in two distinct networks: residual and inverted. Both models are trained on the resultant super-resolution images, whereas the hyperparameters are initialized using Bayesian Optimization (BO). In the testing phase, features are extracted from the self-attention layer and passed to the shallow narrow neural network for classification. The experimental process of the proposed architecture is conducted on three datasets, MLRSNet, Bijie Landslide, and Turkey Earthquake, and improved accuracy of 91.0, 92.8, and 99.4%, respectively. Results are also compared with state-of-the-art (SOTA) techniques and show significant improvement and the model is also evaluated using the lime for the interpretation the outcomes proposed model.