Figure - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Performance comparison on the Beijing Metro flow dataset. We mark the best-performing results by bolded font.
Source publication
Traffic flow forecasting is crucial for improving urban traffic management and reducing resource consumption. Accurate traffic conditions prediction requires capturing the complex spatial-temporal dependencies inherent in traffic data. Traditional spatial-temporal graph modeling methods often rely on fixed road network structures, failing to accoun...
Similar publications
This study proposed a knowledge graph entity extraction and relationship reasoning algorithm based on a graph neural network, using a graph convolutional network and graph attention network to model the complex structure in the knowledge graph. By building an end-to-end joint model, this paper achieves efficient recognition and reasoning of entitie...
With the rapid advancement of quantum computing, the exploration of quantum graph neural networks is gradually emerging. However, the absence of a circuit framework for quantum implementation and limited physical qubits hinder their realization on real quantum computers. To address these challenges, this paper proposes a spatial‐based quantum graph...
In traffic flow prediction tasks, accurately identifying spatio-temporal dependencies with strong correlations to target nodes constitutes a fundamental research challenge. Existing studies reveal that graph convolutional neural networks (GCNs) constrained by conventional adjacency matrices demonstrate limited capabilities in capturing comprehensiv...
Compared to traditional heterogeneous graph (HG) data, Bi-typed Multi-relational Heterogeneous Graph (BMHG) not only have various edge relationships between different types of nodes but also connections among the same type of nodes, which increases modeling complexity. However, existing Heterogeneous Graph Neural Networks (HGNN) primarily focus on...
Citations
... Subsequently, researchers have pivoted their focus from deep learning models that solely focus on graph structure 3,4 to integrating graph structure information, thereby propelling the advancement of graph neural network (GNN)-based methodologies 5 . In recent times, GNN have emerged as leading contenders at the vanguard of deep learning across numerous applications 6 . ...
In the realm of traffic prediction, emerging are methodologies founded on graph convolutional networks. Nonetheless, existing approaches grapple with issues encompassing insufficient sharing patterns, dependence on static relationship presumptions, and an inability to effectively grasp the intricate trends and cyclic attributes of traffic flow. To tackle this challenge, we introduce a novel framework termed Dynamic Graph Convolutional Networks with Temporal Representation Learning for Traffic Flow Prediction (DGCN-TRL). Specifically, a temporal graph convolution block is specifically devised, treating historical time slots as graph nodes and employing graph convolution to process dynamic time series. This approach effectively captures flexible global temporal dependencies, enhancing the model’s ability to comprehend current traffic conditions. Subsequently, a novel dynamic graph constructor is introduced to explore spatial correlations between nodes at specific times and dynamic temporal dependencies across different time points. This meticulous exploration uncovers dynamic spatiotemporal relationships. Finally, a novel temporal representation learning module is developed utilizing a masked subsequence transformer to predict the content of masked subsequences from a fraction of unmasked subsequences and their temporal contexts in a pre-trained manner. This design encourages the model to adeptly learn temporal representations of contextual subsequences from extensive historical data. Empirical evaluations on four real datasets substantiate the superior performance of DGCN-TRL compared to existing methodologies.
... The goal of pose forecasting is to provide accurate predictions of future poses, which can have practical applications in a wide range of fields. For example, in robotics, pose forecasting models enable robots to infer human intentions and predict future movements, facilitating safer, more intuitive collaboration in environments such as manufacturing floors, healthcare, and assistive robotics [1][2][3][4][5][6]. In sports analytics, forecasting player trajectories and body orientations several moments ahead supports tactical decision-making, performance evaluation, and even automated highlight generation. ...
Highlights
This paper presents the GCN-Transformer, a novel deep learning model that integrates Graph Convolutional Networks (GCNs) and Transformers to enhance multi-person pose forecasting. The model effectively captures both spatial and temporal dependencies, improving the performance of pose forecasting. Additionally, a new evaluation metric, Final Joint Position and Trajectory Error (FJPTE), is introduced to provide a more comprehensive assessment of movement dynamics. These contributions establish GCN-Transformer as a state-of-the-art solution in pose forecasting. What are the main findings?We introduce GCN-Transformer, a novel architecture combining Graph Convolutional Networks (GCNs) and Transformers for multi-person pose forecasting.
We propose a new evaluation metric, Final Joint Position and Trajectory Error (FJPTE), which comprehensively assesses both local and global movement dynamics.
What is the implication of the main finding?GCN-Transformer achieves state-of-the-art performances on the CMU-Mocap, MuPoTS- 3D, SoMoF Benchmark, and ExPI datasets, demonstrating superior generalization across different motion scenarios.
The proposed FJPTE metric improves the evaluation of pose forecasting models by accounting for both movement trajectory and final position, enabling better assessments of motion realism.
Abstract
Multi-person pose forecasting involves predicting the future body poses of multiple individuals over time, involving complex movement dynamics and interaction dependencies. Its relevance spans various fields, including computer vision, robotics, human–computer interaction, and surveillance. This task is particularly important in sensor-driven applications, where motion capture systems, including vision-based sensors and IMUs, provide crucial data for analyzing human movement. This paper introduces GCN-Transformer, a novel model for multi-person pose forecasting that leverages the integration of Graph Convolutional Network and Transformer architectures. We integrated novel loss terms during the training phase to enable the model to learn both interaction dependencies and the trajectories of multiple joints simultaneously. Additionally, we propose a novel pose forecasting evaluation metric called Final Joint Position and Trajectory Error (FJPTE), which assesses both local movement dynamics and global movement errors by considering the final position and the trajectory leading up to it, providing a more comprehensive assessment of movement dynamics. Our model uniquely integrates scene-level graph-based encoding and personalized attention-based decoding, introducing a novel architecture for multi-person pose forecasting that achieves state-of-the-art results across four datasets. The model is trained and evaluated on the CMU-Mocap, MuPoTS-3D, SoMoF Benchmark, and ExPI datasets, which are collected using sensor-based motion capture systems, ensuring its applicability in real-world scenarios. Comprehensive evaluations on the CMU-Mocap, MuPoTS-3D, SoMoF Benchmark, and ExPI datasets demonstrate that the proposed GCN-Transformer model consistently outperforms existing state-of-the-art (SOTA) models according to the VIM and MPJPE metrics. Specifically, based on the MPJPE metric, GCN-Transformer shows a 4.7% improvement over the closest SOTA model on CMU-Mocap, 4.3% improvement over the closest SOTA model on MuPoTS-3D, 5% improvement over the closest SOTA model on the SoMoF Benchmark, and a 2.6% improvement over the closest SOTA model on the ExPI dataset. Unlike other models with performances that fluctuate across datasets, GCN-Transformer performs consistently, proving its robustness in multi-person pose forecasting and providing an excellent foundation for the application of GCN-Transformer in different domains.
Traffic flow prediction can guide the rational layout of land use. Accurate traffic flow prediction can provide an important basis for urban expansion planning. This paper introduces a personalized lightweight federated learning framework (PLFL) for traffic flow prediction. This framework has been improved and enhanced to better accommodate traffic flow data. It is capable of collaboratively training a unified global traffic flow prediction model without compromising the privacy of individual datasets. Specifically, a spatiotemporal fusion graph convolutional network (MGTGCN) is established as the initial model for federated learning. Subsequently, a shared parameter mechanism of federated learning is employed for model training. Customized weights are allocated to each client model based on their data features to enhance personalization during this process. In order to improve the communication efficiency of federated learning, dynamic model pruning (DMP) is introduced on the client side to reduce the number of parameters that need to be communicated. Finally, the PLFL framework proposed in this paper is experimentally validated using LPR data from Changsha city. The results demonstrate that the framework can still achieve favorable prediction outcomes even when certain clients lack data. Moreover, the communication efficiency of federated learning under this framework has been enhanced while preserving the distinct characteristics of each client, without significant interference from other clients.
