Asymmetrical Bi-RNN for pedestrian trajectory encoding

Abstract

Pedestrian motion behavior involves a combination of individual goals and social interactions with other agents. In this article, we present an asymmetrical bidirectional recurrent neural network architecture called U-RNN to encode pedestrian trajectories and evaluate its relevance to replace LSTMs for various forecasting models. Experimental results on the Trajnet++ benchmark show that the U-LSTM variant yields better results regarding every available metrics (ADE, FDE, Collision rate) than common trajectory encoders for a variety of approaches and interaction modules, suggesting that the proposed approach is a viable alternative to the de facto sequence encoding RNNs. Our implementation of the asymmetrical Bi-RNNs for the Trajnet++ benchmark is available at: github.com/JosephGesnouin/Asymmetrical-Bi-RNNs-toencode-pedestrian-trajectories

Assessing Cross-dataset Generalization of Pedestrian Crossing Predictors

“…Poses sequences are converted into 2D image-like spatio-temporal representations and selfspatio-temporal attention is applied via CNN-based models for multiple time resolutions. Each remaining feature is independently processed via either U-GRUs [45] or feed forward neural network and fused by either applying temporal and modality attention or sent to a fc layer to predict crossing behaviors.…”

TrouSPI-Net: Spatio-temporal attention on parallel atrous convolutions and U-GRUs for skeletal pedestrian crossing prediction

“…• We secondly represent a skeleton sequence as its evolution of Euclidean pairwise distances of skeletal joints over time and encode them with U-GRUs [36]: a non-symmetrical bidirectional recurrent architecture designed to exploit the bidirectional temporal context and long-term temporal information for challenging skeletal dynamics having similar patterns but different outputs. This compensates for the inabilities of the first stream in learning temporal patterns invariant to locations and viewpoints.…”