Answers are grounded in this paper and cite their sources.
The present study investigates the application of uncertainty modelling for the purpose of detecting pedestrianintentions in contexts pertaining to autonomous driving. The proposed framework integrates two mechanisms: threshold modulation networks for aleatoric uncertainty and cost-sensitive learning for risk-aware decision making.Experiments on the PIE dataset with ResNet50, VGG16, and AlexNet demonstrate that cost-sensitive learning enhances F1-scores marginally (0.05-0.58 percentage points) by prioritising recall for crossing pedestrians. ResNet50 demonstrates the strongest performance (98.30% accuracy, 96.35% F1-score), significantly outperforming more elementary architectures. Threshold networks have been observed to introduce computational overhead, resulting in approximately a doubling of training time, accompanied by slight performance reductions. The study provides empirical evidence for the trade-offs between uncertainty modelling complexity and classification performance in pedestrian intention detection, offering insights for designing safety-oriented perception systems with appropriate computational constraints.nbsp;
No takes yet. Share an insight, caveat, or question.
Li‐Hsuan Huang (2018) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: