Recent open-world representation learning approaches have leveraged CLIP to enable zero-shot 3D object recognition. However, performance on real point clouds with occlusions still falls short due to unrealistic pretraining settings. Additionally, these methods incur high inference costs because they rely on Transformer’s attention modules. In this paper, we make two contributions to address these limitations. First, we propose occlusion-aware text-image-point cloud pretraining to reduce the training-testing domain gap. From 52K synthetic 3D objects, our framework generates nearly 630K partial point clouds for pretraining, consistently improving real-world recognition performances of existing popular 3D networks. Second, to reduce computational requirements, we introduce DuoMamba, a two-stream linear state space model tailored for point clouds. By integrating two spacefilling curves with 1D convolutions, DuoMamba effectively models spatial dependencies between point tokens, offering a powerful alternative to Transformer. When pretrained with our framework, DuoMamba surpasses current state-of-the-art methods while reducing latency and FLOPs, highlighting the potential of our approach for realworld applications. Our code and data are available at ndkhanh360.github.io/project-occtip.

Abstract Online misinformation poses a significant challenge due to its rapid spread and limited supervision. To address this issue, automated rumour detection techniques are essential for countering the negative impact of false information. Previous research primarily focussed on extracting text features, which proved time-consuming and less effective. In this study, we contribute substantially to two domains: rumour detection on Twitter and the evaluation of text embeddings. We thoroughly analyse rumour detection models and compare the quality of text embeddings generated by various fine-tuned BERT-based models. Our findings indicate that our proposed models outperform existing techniques. Notably, when we test these models on combined datasets, we observe significant performance improvements with larger training and testing data sizes. We conclude that carefully considering the dataset, data splitting, and classification techniques is crucial for evaluating solution performance. Additionally, we find that differences in the quality of text embeddings between RoBERTa, BERT, and DistilBERT are insignificant. This challenges existing assumptions and highlights the need for future research to explore these nuances further.

A wide range of respiratory diseases, such as cold and flu, asthma, and COVID-19, affect people’s daily lives worldwide. In medical practice, respiratory sounds are widely used in medical services to diagnose various respiratory illnesses and lung disorders. The traditional diagnosis of such sounds requires specialized knowledge, which can be costly and reliant on human expertise. Recently, cough audio recordings have been used to automate the process of detecting respiratory conditions. This research aims to examine various acoustic features that enhance the performance of machine learning (ML) models in detecting COVID-19 from cough signals. This study investigates the efficacy of three feature extraction techniques, including Mel Frequency Cepstral Coefficients (MFCC), Chroma, and Spectral Contrast features, on two ML algorithms, Support Vector Machine (SVM) and Multilayer Perceptron (MLP), and thus proposes an efficient COVID-19 detection system. The proposed system produces a practical solution and demonstrates higher state-of-the-art classification performance with an AUC of 0.843 on the COUGHVID dataset and 0.953 on the Virufy dataset for COVID-19 detection.

Abstract Containers have emerged recently as a cloud technology for improving and managing cloud resources. They improve resource sharing by allowing instances to run on top of the host’s operating system. Container-based virtualization runs and manages hosted instances via the host kernel. Resource sharing can cause resource contention. In addition, dependent jobs, which may be deployed across multiple hosts, require frequent communication, resulting in a high volume of network traffic and network contention. The majority of existing research focuses on load balancing, with no consideration for the fact that network contention also plays a significant role in container performance. In this research, we propose a Dependency-aware Scheduling algorithm (DAScheduler) that deploys jobs into containers while accounting for both load balancing and job dependencies. The experimental results show that DAScheduler reduces network traffic by more than half and balances the loads. In comparison to one of the existing state-of-the-art techniques, DAScheduler improves overall cloud performance.

Older people are susceptible to fall due to instability in posture and deteriorating health. Immediate access to medical support can greatly reduce repercussions. Hence, there is an increasing interest in automated fall detection, often incorporated into a smart health-care system to provide better monitoring. Existing systems focus on wearable devices that are inconvenient or video monitoring that has privacy concerns. Moreover, these systems provide a limited perspective of their generalization ability as they are tested on datasets containing few activities that have wide disparity in the action space and are easy to differentiate. Complex daily life scenarios pose much greater challenges with activities that overlap in action spaces due to similar posture or motion. To overcome these limitations, we propose a fall detection model, called structure-aware discriminative feature aggregation, based on human skeletons extracted from low-resolution videos. The use of skeleton data ensures privacy and low-resolution videos ensures low hardware and computational cost. Our model captures discriminative structural displacements and motion trends using unified joint and motion features projected onto a shared high-dimensional space. Particularly, the use of separable convolution combined with a powerful graph convolutional network architecture provides improved performance. Extensive experiments on five large-scale datasets with a wide range of evaluation settings show that our model achieves competitive performance with extremely low computational complexity and runs faster than existing models.

Photogrammetry and image processing have been used extensively to examine surface deformations. Digital Image Correlation (DIC) detects the two-dimensional subpixel displacements between two images in order to analyze deformations in geomechanical structures. In this study, the case of large deformations involving nonlinear elastic behavior of the material has been examined with the use of a physical model of Ethylene Vinyl Acetate (EVA) foam subject to axial strain up to 15%. The associated strain and displacement fields are reconstructed using DIC and compared with the Finite Element Method. DIC is used to analyze the critical hysteresis effect of the material which occurs during loading and unloading process. The results show that DIC is a reliable technique to analyze the nonlinear elastic behavior of materials.

Digital image correlation (DIC) is a well-known contact-less technique offering highly accurate full-field deformation measurement using grayscale images. The practical implementation of DIC is still facing many challenges, especially limitations of accuracy in measuring small displacement gradients for solids in geosciences and biomedi-cal engineering. In this paper, we introduce a novel approach in which color images are employed to enhance the performance of DIC. A complete framework for Color DIC has been proposed and tested. The results show that Color DIC performs significantly better than grayscale DIC for measurement of small strains by a factor of 2.

Reconstruction of displacement and strain fields in geomechanical structures from surface images is a challenging task. Digital Image Correlation (DIC) is a well known technique to achieve these tasks if deformation is continuous but it fails in the presence of discontinuities. This paper investigates the application of the DIC technique to displacement and strain field reconstruction in the presence of discontinuities, and presents a post-processing algorithm that leverages the convergence results in DIC to reconstruct displacement and strain fields around discontinuities with high accuracy. The proposed algorithm uses the results obtained from DIC and concentrates on the area where DIC fails. Pattern matching is conducted on the area around the discontinuities and associated displacement is found for each pixel. The proposed algorithm is tested using two different discontinuity scenarios: dislocation and fracture in structures. The results show that the proposed algorithm successfully reconstructs the displacement and strain fields to sub pixel accuracy of 1/10th of a pixel.