The fusion of visual and tactile senses allows robots to reconstruct and understand objects, aiding in downstream tasks such as object recognition and object grasping. However, most existing visual-tactile reconstruct...
详细信息
Outlier detection is one of the hot topics in the field of machine learning and data mining. At present, there are many kinds of outlier detection algorithms. The accuracies of traditional outlier detection algorithms...
详细信息
Recent preliminary research has examined cross-modal attacks, enabling the transition of unlabeled attacks from images to videos. The current cross-modal attacks are mainly dense attacks, while sparse attacks have not...
详细信息
In the era of open cloud, cloud API is an important component and key enabling technology to achieve efficient data transmission, artificial intelligence algorithm empowerment, and software development cost reduction ...
详细信息
Due to data imbalance, existing spammer group detection methods often yield suboptimal performance. Moreover, many of these approaches operate as black boxes, offering little to no interpretability for their detection...
详细信息
Currently, many researchers aim to achieve automatic depression level prediction via speech and video behavior analysis. However, previous works have struggled to decompose audio and video sequences into the informati...
详细信息
Novel view synthesis (NVS) from limited observations continues to be an important and persistent challenge. Currently, methods based on Neural Radiance Fields (NeRF) are not very efficient, and although methods based ...
Novel view synthesis (NVS) from limited observations continues to be an important and persistent challenge. Currently, methods based on Neural Radiance Fields (NeRF) are not very efficient, and although methods based on 3D Gaussian Splatting (3DGS) outperform NeRF in terms of rendering quality and speed, they still lack sufficient details. To tackle these issues, we introduce a new method for densifying the point cloud, which enhances the rendering effect of 3DGS-based techniques under sparse view conditions. First, we introduce a mask-based densification technique to improve rendering details under limited input views. Second, We propose a monocular pixel depth-based mapping method that leverages a pre-trained model to predict depth, effectively normalizing point locations within the resulting point cloud. Lastly, we implement a filtering method based on depth and RGB color to minimize noise introduced by the additional data. We conduct comparative experiments on the LLFF, Mip-NeRF 360, and Blender datasets, demonstrating that our method outperforms existing approaches in both evaluation metrics and visual quality, thereby validating its effectiveness.
Multi-access Edge Computing (MEC) has been a promising solution that enables Internet of Things (IoT) devices to support computation-intensive applications by offloading some tasks to the network edge. However, most e...
详细信息
In this paper, we focus on efficient processing of XML keyword queries based on smallest lowest common ancestor (SLCA) semantics. For a given query Q with m keywords, we propose to use stable matches as the basis fo...
详细信息
In this paper, we focus on efficient processing of XML keyword queries based on smallest lowest common ancestor (SLCA) semantics. For a given query Q with m keywords, we propose to use stable matches as the basis for SLCA computation, where each stable match M consists of m nodes that belong to the m distinct keyword inverted lists of Q. M satisfies that no other lowest common ancestor (LCA) node of Q can be found to be located after the first node of M and be a descendant of the LCA of M, based on which the operation of locating a stable match can skip more useless nodes. We propose two stable match based algorithms for SLCA computation, i.e., BSLCA and HSLCA. BSLCA processes two keyword inverted lists each time from the shortest to the longest, while HSLCA processes all keyword inverted lists in a holistic way to avoid the problem of redundant computation invoked by BSLCA. Our extensive experimental results verify the performance advantages of our methods according to various evaluation metrics.
In this paper, we try to systematically study how to perform doctor recommendation in medical social net- works (MSNs). Specifically, employing a real-world medical dataset as the source in our work, we propose iBol...
详细信息
In this paper, we try to systematically study how to perform doctor recommendation in medical social net- works (MSNs). Specifically, employing a real-world medical dataset as the source in our work, we propose iBole, a novel hybrid multi-layer architecture, to solve this problem. First, we mine doctor-patient relationships/ties via a time-constraint probability factor graph model (TPFG). Second, we extract network features for ranking nodes. Finally, we propose RWR- Model, a doctor recommendation model via the random walk with restart method. Our real-world experiments validate the effectiveness of the proposed methods. Experimental results show that we obtain good accuracy in mining doctor-patient relationships from the network, and the doctor recommendation performance is better than that of the baseline algorithms: traditional Ranking SVM (RSVM) and the individual doctor recommendation model (IDR-Model). The results of our RWR-Model are more reasonable and satisfactory than those of the baseline approaches.
暂无评论