PIT: Progressive Interaction Transformer for Pedestrian Crossing Intention Prediction

Zhou, Yuchen; Tan, Guang; Zhong, Rui; Li, Yaokun; Gou, Chao

doi:10.1109/tits.2023.3309309

Cited by 19 publications

(3 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Table II shows the model comparison results for the PIE and JAAD dataset. The PedAST-GCN model achieves comparable or better results compared to the state-of-the-art Pedestrian Graph + model [19] and PIT [26]. The improvement is attributed to the use of multi-modality data, such as pose keypoints data to capture pedestrian pose and action information, bounding box data to capture pedestrian motion, distance, size information, and vehicle speed data to capture vehicle movement information.…”

Section: Resultsmentioning

confidence: 98%

“…Yang et al [13] developed a hierarchical RNN-based model and incorporated the global context (semantic map) in the PCPA model to capture the scene information, and compared different combinations of data type streams. Zhou et al [26] introduced a transformer-based model for pedestrian crossing intention prediction, which incorporates a temporal fusion block and a self-attention mechanism to capture richer information. Although these models are reported to achieve a high prediction accuracy, they use complex models for preparing different types of model input data which tends to be computationally intensive and limited for real-time applications.…”

Section: A Pedestrian Crossing Intention Predictionmentioning

confidence: 99%

“…• PIT [26]. It uses the pose keypoints, image crop, vehicle speed, bounding box, and global image as the input and the transformer-based model for pedestrian crossing intention prediction.…”

Section: B Compared Models and Evaluation Metricsmentioning

confidence: 99%

See 2 more Smart Citations

PedAST-GCN: Fast Pedestrian Crossing Intention Prediction Using Spatial–Temporal Attention Graph Convolution Networks

Ling,

Ma,

Zhang

et al. 2024

IEEE Trans. Intell. Transport. Syst.

View full text Add to dashboard Cite

Accurately and timely predicting pedestrian crossing intentions in real-time is critical for operating intelligent vehicles on roads. Although existing models achieve promising accuracy using complex models and video image data, they are constrained for real-time practical use given the high model complexity, time-consuming data preprocessing, and low-quality image data in the wild. To address these, the paper proposes a Spatial-Temporal Attention Graph Convolution Network model for fast pedestrian crossing intention prediction (PedAST-GCN). It uses a lightweight GCN model as the backbone network with simple but robust graph representations of pedestrian crossing intention modality features, including pedestrian pose, bounding box, and vehicle speeds. The model is validated by comparing it with state-of-the-art models on two large-scale public datasets (JAAD and PIE). The results highlight the better performance of the PedAST-GCN model for pedestrian crossing intention prediction in terms of accuracy and computation times. The ablation analysis confirms the value of the backbone layer and graph design, the designed modality features, the effectiveness of attention mechanisms in capturing long-term dependencies (spatial-temporal attention) and fusing heterogeneous features (modality attention), and the robust performance across various observation lengths and in the presence of noisy data.

show abstract

Section: Resultsmentioning

confidence: 98%

Section: A Pedestrian Crossing Intention Predictionmentioning

confidence: 99%