实验的Action Genome (AG) 数据集是斯坦福大学在 Charades 数据集(见ECCV'16论文“Hollywood in homes: Crowdsourcing data collection for activity understanding“)之上提供帧级场景图标签(详细看CVPR‘20论文“Action genome: Actions as compositions of spatio-temporal scene graphs“)。它包括35 个目标类(没有人...
4. Beyond Spatio-Temporal Representations: Evolving Fourier Transform for Temporal Graphs 5. GeoLLM: Extracting Geospatial Knowledge from Large Language Models 6. A Generative Pre-Training Framework for Spatio-Temporal Graph Transfer Learning 7. AirPhyNet: Harnessing Physics-Guided Neural Networks for Ai...
as space-time region graphs》2018 --- 视频分类 论文:它们忽略了帧内或帧间不同身体部位之间的关系,是基于图像的,不考虑时间关系。 《Adaptive graph representation learning for video person re-identification》2019---引入图神经网络,利用姿态对齐连接和特征相似性连接实现相关区域特征之间的上下文交互。此外,该...
32 proposed the MS-AAGCN model that uses a data-driven approach to increase its flexibility and generalization capabilities; the authors confirmed that the adaptive learning graph topology is more suitable for action recognition tasks than human-based graphs. The above approaches are all valid spatio...
5. [Oral] GraFITi: Graphs for Forecasting Irregularly Sampled Time Series 作者：Yalavarthi, Vijaya Krishna*; Madhusudhanan, Kiran; Scholz, Randolf; Ahmed, Nourhan; Burchert, Johannes; Jawed, Shayan; Born, Stefan; Schmidt-Thieme, Lars 关键词：不规则时间序列，图（Graph）arXiv链接：...
The code and models arepublicly available at https://github.com/soumbane/STSGT.KEYWORDSspatial-temporal graphs, COVID-19 forecasting, transformers1 INTRODUCTIONThe Coronavirus disease 2019 (COVID-19) is caused by severe acuterespiratory syndrome coronavirus 2 (SARS-CoV-2). The f i rst known...
4. Beyond Spatio-Temporal Representations: Evolving Fourier Transform for Temporal Graphs 链接:https://openreview.net/forum?id=uvFhCUPjtI 分数:5, 3, 8, 6 关键词:时空表示,傅里叶变换 TL; DR:第一篇提出将演化时间图变换到频域概念的工作,称之为“演化图傅里叶变换(EFT) ...
Our code is publicly available at this https URLdoi:10.48550/arXiv.2207.07783Min, KyleRoy, SouryaTripathi, SubarnaGuha, TanayaMajumdar, SomdebSpringer, ChamEuropean Conference on Computer Vision
然而,正如前面提到的,骨架是图(graphs)的形式,而不是2D或3D网格,这使得使用卷积网络等经过验证的模型变得困难。最近,图神经网络(Graph Neural networks, GCNs)将卷积神经网络(convolutional Neural networks, CNNs)推广到任意结构的图形,受到越来越多的关注,并成功地应用于许多应用中,如图像分类(Bruna et al. 2014...
5. [Oral] GraFITi: Graphs for Forecasting Irregularly Sampled Time Series 6. IVP-VAE: Modeling EHR Time Series with Initial Value Problem Solvers 7. Cross-Domain Contrastive Learning for Time Series Clustering 8. SimPSI: A Simple Strategy to Preserve Spectral Information in Time Series Data...