A transformer model that fuses predicted pedestrian trajectories and vehicle speed with scene images achieves competitive accuracy and the lowest inference time on pedestrian crossing intention benchmarks.
Visual attention network,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
TrajFusionNet: Pedestrian Crossing Intention Prediction via Fusion of Sequential and Visual Trajectory Representations
A transformer model that fuses predicted pedestrian trajectories and vehicle speed with scene images achieves competitive accuracy and the lowest inference time on pedestrian crossing intention benchmarks.