TY  - EJOU
AU  - Landry, François G. 
AU  - Akhloufi, Moulay A. 

TI  - TrajFusionNet: Pedestrian Crossing Intention Prediction via Fusion of Sequential and Visual Trajectory Representations
T2  - Computer Modeling in Engineering \& Sciences

PY  - 
VL  - 
IS  - 
SN  - 1526-1506

AB  - With the introduction of vehicles with autonomous capabilities on public roads, predicting pedestrian crossing intention has emerged as an active area of research. The task of predicting pedestrian crossing intention involves determining whether pedestrians in the scene are likely to cross the road or not. In this work, we propose TrajFusionNet, a novel transformer-based model that leverages future pedestrian trajectory and vehicle speed predictions as priors for predicting crossing intention. TrajFusionNet comprises two branches: a Sequence Attention Module (SAM) and a Visual Attention Module (VAM). The SAM branch learns from a sequential representation of the observed and predicted pedestrian trajectory and vehicle speed. Complementarily, the VAM branch learns from a visual representation of the observed and predicted pedestrian trajectory by overlaying corresponding pedestrian bounding boxes onto scene images. In terms of performance, TrajFusionNet achieves state-of-the-art results on the Joint Attention in Autonomous Driving (JAAD) and Pedestrian Intention Estimation (PIE) datasets. By utilizing a small number of lightweight modalities, it also achieves the lowest total inference time (including model runtime and data preprocessing) among state-of-the-art approaches.
KW  - Pedestrian crossing intention; pedestrian trajectory; autonomous vehicle; transformer

DO  - 10.32604/cmes.2026.086105