SGCN:Sparse Graph Convolution Network for Pedestrian Trajectory Prediction

Shi, Liushuai; Wang, Le; Long, Chengjiang; Zhou, Sanping; Zhou, Mo; Niu, Zhenxing; Hua, Gang

doi:10.1109/cvpr46437.2021.00888

Cited by 141 publications

(123 citation statements)

References 23 publications

Supporting

Mentioning

123

Contrasting

Order By: Relevance

“…For each time step, the predicted location with the highest probability is used to calculate the ADE and FDE (see Methods). GIFNet outperforms the state-of-the-art machinelearning-based trajectory prediction methods (Sophie 22 , STGAT 34 , SGAN 35 , Social-STGCNN 20 , SGCN 36 ), and the baseline method 'Linear'. Among these methods, only our GIFNet encodes all four kinds of features: trajectory, visual orientation, neighbour trajectory, and group interaction state.…”

Section: Predicting the Proxemics Fieldmentioning

confidence: 98%

Group interaction field for learning and explaining pedestrian anticipation

Wang

Chen

Puhua

et al. 2022

Preprint

View full text Add to dashboard Cite

Anticipating others’ actions is innate and essential for humans to navigate and interact well with others in dense crowds. Such ability is urgently demanded by unmanned systems, such as service robots, self-driving cars, etc. However, existing solutions hardly predict pedestrian anticipation accurately because the influence of group-related social behaviours has not been well considered. While the group and group interactions are ubiquitous and significantly influence pedestrian anticipation, this influence is diverse and subtle, making it difficult to explicitly quantify. We innovate the group interaction field (GIF), a novel group-aware representation that quantifies the pedestrian anticipation into the probability field of pedestrians' future locations and attention orientations. An end-to-end neural network, GIFNet, is proposed to estimate the GIF from explicit multi-dimensional observations. GIFNet quantifies group behaviours' influence by formulating a group interaction graph with propagation and attention adaptive to group size and dynamic interaction states. The experimental results show that the GIF can effectively represent the change of pedestrians’ anticipation under the prominent impact of group behaviours, and predict pedestrians’ future states accurately. Moreover, the GIF contributes to explaining various anticipations of pedestrians’ behaviour in different social states. The proposed GIF can eventually innovate unmanned systems to work in a human-like manner, comply with social norms, and thereby promote a harmonious human–machine relationship.

show abstract

Section: Predicting the Proxemics Fieldmentioning

confidence: 98%

Group interaction field for learning and explaining pedestrian anticipation

Wang

Chen

Puhua

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Furthermore, we highlight the difficulties that intention-based models encounter in highly socially environments experimenting on a synthetic dataset where predictions must consider social interaction rules. Finally, an effective way of modeling social interactions, which is gaining an increasing attention, is to model agents as nodes in a graph and then process it with a Graph Neural Network (GNN) [9], [33], [34].…”

Section: Trajectory Predictionmentioning

confidence: 99%

SMEMO: Social Memory for Trajectory Forecasting

Marchetti¹,

Becattini²,

Seidenari³

et al. 2022

Preprint

View full text Add to dashboard Cite

Effective modeling of human interactions is of utmost importance when forecasting behaviors such as future trajectories. Each individual, with its motion, influences surrounding agents since everyone obeys to social non-written rules such as collision avoidance or group following. In this paper we model such interactions, which constantly evolve through time, by looking at the problem from an algorithmic point of view, i.e. as a data manipulation task. We present a neural network based on an end-to-end trainable working memory, which acts as an external storage where information about each agent can be continuously written, updated and recalled. We show that our method is capable of learning explainable cause-effect relationships between motions of different agents, obtaining state-of-the-art results on multiple trajectory forecasting datasets.

show abstract

“…Thanks to the representational power of deep neural networks, trajectory prediction is recently dominated by deep learning based methods, such as Recurrent Neural Networks (RNNs) [3], Generative Adversarial Networks (GANs) [29], Graph Convolutional Networks (GCNs) [66,84] and Transformers [106]. S-LSTM [3] aggregates the interaction information through a pooling mechanism.…”

Section: Homogeneous Trajectory Predictionmentioning

confidence: 99%

Unlimited Neighborhood Interaction for Heterogeneous Trajectory Prediction

Zheng¹,

Wang

Zhou

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Understanding complex social interactions among agents is a key challenge for trajectory prediction. Most existing methods consider the interactions between pairwise traffic agents or in a local area, while the nature of interactions is unlimited, involving an uncertain number of agents and non-local areas simultaneously. Besides, they treat heterogeneous traffic agents the same, namely those among agents of different categories, while neglecting people's diverse reaction patterns toward traffic agents in different categories. To address these problems, we propose a simple yet effective Unlimited Neighborhood Interaction Network (UNIN), which predicts trajectories of heterogeneous agents in multiple categories. Specifically, the proposed unlimited neighborhood interaction module generates the fused-features of all agents involved in an interaction simultaneously, which is adaptive to any number of agents and any range of interaction area. Meanwhile, a hierarchical graph attention module is proposed to obtain category-to-category interaction and agent-to-agent interaction. Finally, parameters of a Gaussian Mixture Model are estimated for generating the future trajectories. Extensive experimental results on benchmark datasets demonstrate a significant performance improvement of our method over the state-of-the-art methods.

show abstract

SGCN:Sparse Graph Convolution Network for Pedestrian Trajectory Prediction

Cited by 141 publications

References 23 publications

Group interaction field for learning and explaining pedestrian anticipation

Group interaction field for learning and explaining pedestrian anticipation

SMEMO: Social Memory for Trajectory Forecasting

Unlimited Neighborhood Interaction for Heterogeneous Trajectory Prediction

Contact Info

Product

Resources

About