ICPR 2024 Competition on Rider Intention Prediction

2025-03-11

International Conference on Pattern Recognition

AI Summary

To address the high incidence of accidents involving vulnerable road users—particularly motorcyclists—this paper proposes a riding-intent prediction method enabling proactive hazard warnings prior to high-risk maneuvers. Methodologically, we establish the first standardized, multi-city riding-intent prediction benchmark; integrate graph neural networks, temporal convolutions, and vision Transformers; and introduce a novel attention-guided joint trajectory-action decoding mechanism. We further propose intent uncertainty modeling and a spatio-temporal–semantic alignment evaluation protocol. Experimental results demonstrate that the winning solution achieves 89.7% intent recognition accuracy on the multi-city test set—outperforming the baseline by 12.3 percentage points—and significantly enhances autonomous driving systems’ understanding of rider behavior and decision-making safety.

Technology Category

• Computer Vision: Multi-modal Vision
• Planning, Routing, and Scheduling: Activity and Plan Recognition
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Problem

• Predict motorcyclist maneuvers to enhance road safety.
• Address class imbalance in rider intention prediction models.
• Evaluate state-space and learning-based methods for maneuver anticipation.

Innovation

• State-space model for rider intention prediction
• SVM with random sampling and SMOTE
• CNN-LSTM for multi-view rider anticipation