Background: Criminal actions using falsified papers are on the rise. Any changes to the document's content are usually made with ballpoint and gel pens. All changes made to a document using a pen necessitate ink a...
详细信息
Aiming at the problems of high randomness and low prediction accuracy of power load data, a power load prediction model is formed by integrating Singular Spectrum Analysis(SSA) and a Gated Recurrent Unit(GRU) network ...
详细信息
Entity alignment is the task of identifying entities from different knowledge graphs (KGs) that point to the same item and is important for KG fusion. In the real world, due to the heterogeneity between different KGs,...
详细信息
Entity alignment is the task of identifying entities from different knowledge graphs (KGs) that point to the same item and is important for KG fusion. In the real world, due to the heterogeneity between different KGs, equivalent entities often have different relations around them, so it is difficult for Graph Convolutional Network (GCN) to accurately learn the relation information in the KGs. Moreover, to solve the problem regarding inadequate utilisation of relation information in entity alignment, a novel GCN-based model, joint Unsupervised Relation Alignment for Entity Alignment (URAEA), is proposed. The model first employs a novel method for calculating relation embeddings by using entity embeddings, then constructs unsupervised seed relation alignments through these relation embeddings, and finally performs entity alignment together with relation alignment. In addition, the seed entity alignments are expanded based on the generated seed relation alignments. Experiments conducted on three real-world datasets show that this approach outperforms state-of-the-art methods.
Semi -supervised learning has achieved extraordinary success in prevalent image -classification benchmarks. However, a class -balanced distribution that differs notably from real -world data distribution is required. ...
详细信息
Semi -supervised learning has achieved extraordinary success in prevalent image -classification benchmarks. However, a class -balanced distribution that differs notably from real -world data distribution is required. In general, models trained under class-imbalanced semi -supervised learning conditions are severely biased towards the majority classes. To address this issue, we propose a novel framework called ABAE by implanting an Auxiliary Balanced AutoEncoder branch into existing semi -supervised learning algorithms. Considering that adaptive feature augmentation for different classes can alleviate confirmation bias, we devise a class -aware reconstruction loss to train the AutoEncoder module. To further smooth the output, we adopt a graph -based label propagation scheme at the end of the AutoEncoder. Extensive experiments on CIFAR-10/100-LT, SVHN-LT and Small demonstrate the effectiveness of ABAE.
In professional vocal training, a way to evaluate the quality of vocalization is often needed. In order to solve various problems caused by the lack of professional instructors, a vocal training system for detecting t...
详细信息
Masked Image Modeling (MIM), following "mask-and-reconstruct" scheme, is a promising self-supervised method to learn scalable visual representation. Studies indicate that selecting an effective mask strategy...
详细信息
In recent years, mobile applications have attempted to significantly improve our lives. Android and iOS are the two most popular mobile platforms, yet they have varied settings (OS versions, screens, manufacturers, et...
详细信息
Addressing the challenge of effectively capturing features in contemporary video tasks, we propose an action recognition approach grounded in keyframe filtering and feature fusion. Our method comprises two core module...
详细信息
ISBN:
(纸本)9789819785100;9789819785117
Addressing the challenge of effectively capturing features in contemporary video tasks, we propose an action recognition approach grounded in keyframe filtering and feature fusion. Our method comprises two core modules. The keyframe screening module employs an attention mechanism to segregate the input depth feature map sequence into two distinct tensors, effectively reducing spatial redundancy computation and enhancing key feature capture. The other spatio-temporal and action feature module features two branches with divergent structures, performing spatio-temporal and action feature extraction on the differentiated features from the previous module. Through these closely linked modules, our approach effectively discerns and extracts meaningful video features for subsequent classification tasks. We construct an end-to-end deep learning model using established frameworks, training and validating it on a generic video dataset, and confirm its efficacy through comparison and ablation experiments. Experiments conducted on this dataset demonstrate that our model surpasses the majority of prior works.
We present a novel approach for 3D sketching in Augmented Reality (AR). Unlike most existing methods that rely on positioning devices or algorithms, our approach only needs a mobile device to realize 3D sketching in A...
详细信息
Multi-view stereo (MVS) methods based on deep learning have developed rapidly in recent years, but inaccuracies in reconstruction due to the general effect of feature extraction and poor correlation between cost volum...
详细信息
Multi-view stereo (MVS) methods based on deep learning have developed rapidly in recent years, but inaccuracies in reconstruction due to the general effect of feature extraction and poor correlation between cost volumes are still present, opening possibilities for improvement in reconstruction accuracy and completeness. We therefore develop a hierarchical MVS network model with cost volume separation and fusion to mitigate these problems. First, to obtain a more complete and accurate feature information from the input images, a U-shape feature extraction module was designed that outputs feature information simultaneously according to a hierarchical structure composed of three different scales. Then, to enhance the learning ability of the network structure for features, we introduced attention mechanisms to the extracted features that focus on and learn the highlighted features. Finally, in the cost volume regularization stage, a cost volume separation and fusion module was designed in the structure of a hierarchical cascade. This module separates the information within the small-scale cost volume, passes it to the lower level cost volume for fusion, and performs a coarse-to-fine depth map estimation. This model results in substantial improvements in reconstruction accuracy and completeness. The results of extensive experiments on the DTU dataset show that our method performs better than Cascade-MVSNet by about 10.2% in accuracy error (acc.), 7.6% in completeness error (comp.), and 9.0% in overall error (overall), with similar performance in the reconstruction completeness, showing the validity of our module.
暂无评论