Rapid damage identification and classification in disastrous situations and natural disasters are crucial for efficiently directing aid and resources. With the development of deep learning techniques and the availability of imagery content on social media platforms, extensive research has focused on damage assessment. Through the use of geospatial data related to such incidents, the visual characteristics of these images can quickly determine the safety situation in the region. However, training accurate disaster classification models has proven to be challenging due to the lack of labeled imagery data in this domain. This paper proposes a disaster classification framework, which combines a set of synthesized diverse disaster images generated using generative adversarial networks (GANs) and domain-specific fine-tuning of a deep convolutional neural network (CNN)-based model. The proposed model utilizes bootstrap aggregating (bagging) to further stabilize the target predictions. Since past work in this domain mainly suffers from limited data resources, a sample dataset that highlights the issue of imbalanced classification of multiple natural disasters was constructed and augmented. Qualitative and quantitative experiments show the validity of the data augmentation method employed in producing a balanced dataset. Further experiments with various evaluation metrics verified the proposed framework’s accuracy and generalization ability across different classes for the task of disaster classification in comparison to other state-of-the-art techniques. Furthermore, the framework outperforms the other models by an average validation accuracy of 11%. These results provide a deep learning solution for real-time disaster monitoring systems to mitigate the loss of lives and properties.

Emotion recognition is crucial in artificial intelligence, particularly in the domain of human–computer interaction. The ability to accurately discern and interpret emotions plays a critical role in helping machines to effectively decipher users’ underlying intentions, allowing for a more streamlined interaction process that invariably translates into an elevated user experience. The recent increase in social media usage, as well as the availability of an immense amount of unstructured data, has resulted in a significant demand for the deployment of automated emotion recognition systems. Artificial intelligence (AI) techniques have emerged as a powerful solution to this pressing concern in this context. In particular, the incorporation of multimodal AI-driven approaches for emotion recognition has proven beneficial in capturing the intricate interplay of diverse human expression cues that manifest across multiple modalities. The current study aims to develop an effective multimodal emotion recognition system known as MM-EMOR in order to improve the efficacy of emotion recognition efforts focused on audio and text modalities. The use of Mel spectrogram features, Chromagram features, and the Mobilenet Convolutional Neural Network (CNN) for processing audio data are central to the operation of this system, while an attention-based Roberta model caters to the text modality. The methodology of this study is based on an exhaustive evaluation of this approach across three different datasets. Notably, the empirical findings show that MM-EMOR outperforms competing models across the same datasets. This performance boost is noticeable, with accuracy gains of an impressive 7% on one dataset and a substantial 8% on another. Most significantly, the observed increase in accuracy for the final dataset was an astounding 18%.

One of the primary tasks undertaken by autonomous vehicles (AVs) is object detection, which comes ahead of object tracking, trajectory estimation, and collision avoidance. Vulnerable road objects (e.g., pedestrians, cyclists, etc.) pose a greater challenge to the reliability of object detection operations due to their continuously changing behavior. The majority of commercially available AVs, and research into them, depends on employing expensive sensors. However, this hinders the development of further research on the operations of AVs. In this paper, therefore, we focus on the use of a lower-cost single-beam LiDAR in addition to a monocular camera to achieve multiple 3D vulnerable object detection in real driving scenarios, all the while maintaining real-time performance. This research also addresses the problems faced during object detection, such as the complex interaction between objects where occlusion and truncation occur, and the dynamic changes in the perspective and scale of bounding boxes. The video-processing module works upon a deep-learning detector (YOLOv3), while the LiDAR measurements are pre-processed and grouped into clusters. The output of the proposed system is objects classification and localization by having bounding boxes accompanied by a third depth dimension acquired by the LiDAR. Real-time tests show that the system can efficiently detect the 3D location of vulnerable objects in real-time scenarios.

In autonomous driving, object detection is considered a base step to many subsequent processes. However, object detection is challenged by loss in visibility caused by rain. Rainfall occurs in two main forms, which are streaks and streaks accumulations. Each degradation type imposes different effect on the captured videos; therefore, they cannot be mitigated in the same way. We propose a lightweight network which mitigates both types of rain degradation in real-time, without negatively affecting the object-detection task. The proposed network consists of two different modules which are used progressively. The first one is a progressive ResNet for rain streaks removal, while the second one is a transmission-guided lightweight network for rain streak accumulation removal. The network has been tested on synthetic and real rainy datasets and has been compared with state-of-the-art (SOTA) networks. Additionally, time performance evaluation has been performed to ensure real-time performance. Finally, the effect of the developed deraining network has been tested on YOLO object-detection network. The proposed network exceeded SOTA by 1.12 dB in PSNR on the average result of multiple synthetic datasets with 2.29× speedup. Finally, it can be observed that the inclusion of different lightweight stages works favorably for real-time applications and could be updated to mitigate different degradation factors such as snow and sun blare.

Ant colony optimization algorithm (ACO) is a soft computing metaheuristic that belongs to swarm intelligence methods. ACO has proven a well performance in solving certain NP-hard problems in polynomial time. This paper proposes the analysis, design and implementation of ACO as a parallel metaheuristics using the OpenMP framework. To improve the efficiency of ACO parallelization, different related aspects are examined, including scheduling of threads, race hazards and efficient tuning of the effective number of threads. A case study of solving the traveling salesman problem (TSP) using different configurations is presented to evaluate the performance of the proposed approach. Experimental results show a significant speedup in execution time for more than 3 times over the sequential implementation.