Cross-domain Trajectory Prediction with CTP-Net

TL;DR

This paper introduces CTP-Net, a domain adaptation framework for pedestrian trajectory prediction that aligns features across domains and regularizes future predictions to improve cross-domain generalization.

Contribution

The paper proposes a novel CTP-Net architecture that uses adversarial feature alignment and trajectory consistency regularization for effective domain adaptation in trajectory prediction.

Findings

Significant improvement in cross-domain pedestrian trajectory prediction accuracy.

Effective feature alignment between source and target domains.

Enhanced trajectory prediction consistency across domains.

Abstract

Most pedestrian trajectory prediction methods rely on a huge amount of trajectories annotation, which is time-consuming and expensive. Moreover, a well-trained model may not effectively generalize to a new scenario captured by another camera. Therefore, it is desirable to adapt the model trained on an annotated source domain to the target domain. To achieve domain adaptation for trajectory prediction, we propose a Cross-domain Trajectory Prediction Network (CTP-Net). In this framework, encoders are used in both domains to encode the observed trajectories, then their features are aligned by a cross-domain feature discriminator. Further, considering the consistency between the observed and the predicted trajectories, a target domain offset discriminator is utilized to adversarially regularize the future trajectory predictions to be in line with the observed trajectories. Extensive experiments demonstrate the effectiveness of our method on domain adaptation for pedestrian trajectory prediction.