ContextMotionCLIP
Find every recording of “a cyclist turning left without signalling” by typing it, instead of scrolling through hours of logged drives.
What it does
Rare behaviour is what breaks automated vehicles, and rare behaviour is exactly what is hard to find in a dataset. Text-to-motion retrieval makes the long tail reachable for training and for evaluation.
Abstract
Autonomous driving systems must operate reliably in safety-critical scenarios, particularly those involving unusual or complex behavior by Vulnerable Road Users (VRUs). Identifying these edge cases in driving datasets is essential for robust evaluation and generalization, but retrieving such rare human behavior scenarios within the long tail of large-scale datasets is challenging. To support targeted evaluation of autonomous driving systems in diverse, human-centered scenarios, we propose a novel context-aware motion retrieval framework. Our method combines Skinned Multi-Person Linear (SMPL)-based motion sequences and corresponding video frames before encoding them into a shared multimodal embedding space aligned with natural language. Our approach enables the scalable retrieval of human behavior and their context through text queries. This work also introduces our dataset WayMoCo, an extension of the Waymo Open Dataset. It contains automatically labeled motion and scene context descriptions derived from generated pseudo-ground-truth SMPL sequences and corresponding image data. Our approach outperforms state-of-the-art models by up to 27.5% accuracy in motion-context retrieval, when evaluated on the WayMoCo dataset.
Result
Citation
@article{Englmeier2025Contextbased,
title = {Context-based Motion Retrieval using Open Vocabulary Methods for Autonomous Driving},
author = {Stefan Englmeier and Max A. Büttner and Katharina Winter and Fabian B. Flohr},
url = {https://arxiv.org/abs/2508.00589},
year = {2025},
date = {2025-01-01},
journal = {arXiv preprint arXiv:2508.00589},
keywords = {},
pubstate = {published},
tppubtype = {article}
}