TraVEL: Trajectory-Guided Video Embedding Learning for Driving-Video Retrieval

• URL: http://arxiv.org/abs/2608.13495v1
• Date: Thu, 13 Aug 2026 17:24:23 GMT
• Title: TraVEL: Trajectory-Guided Video Embedding Learning for Driving-Video Retrieval
• Abstract summary: Efficiently retrieving relevant clips from driving logs is essential for data curation, model development, and safety analysis.<n>General-purpose embedding models often rely on shortcuts from static scene context.<n>We introduce TraVEL, a motion-aware fine-tuning framework that uses ego-trajectory similarity as a reward.
• Score: 15.588457512574832
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Efficiently retrieving relevant clips from large-scale driving logs is essential for data curation, model development, and safety analysis. Structured and rule-based retrieval systems can explicitly target driving events, but typically require expert-defined rules, auxiliary data, and multi-stage perception pipelines. Multimodal embedding models offer a simpler and more efficient alternative by representing each video with a single searchable vector. However, general-purpose models often rely on shortcuts from static scene context and struggle to distinguish motion-centric events, such as turning left versus right or accelerating versus decelerating. In this work, we study how to adapt a general-purpose multimodal embedding model to driving-video retrieval. We first fine-tune Qwen3-VL-Embedding on paired clips and reasoning traces from nuReasoning using an InfoNCE objective. While this stage substantially improves overall retrieval, caption supervision alone remains insufficient for fine-grained motion understanding. We therefore introduce TraVEL (Trajectory-Guided Video Embedding Learning), a motion-aware fine-tuning framework that uses ego-trajectory similarity as a reward within Group Relative Policy Optimization. Trajectories serve only as privileged training supervision; retrieval still operates on single-vector video embeddings without ego poses, expert rules, or auxiliary perception outputs. We further construct a driving-video retrieval benchmark from nuReasoning. Experiments show that TraVEL improves motion-centric retrieval across model scales: relative to SFT, it raises longitudinal and lateral mAP by 9.8 and 4.7 points at 2B, with corresponding gains of 7.2 and 1.5 points at 8B. TraVEL thus combines physically grounded supervision with efficient embedding-based search.

Related papers

• VideoScout: Learning Agentic Active Exploration with Adaptive Reasoning Pacing for Long Video Understanding [61.00809248898572]
  Long video understanding is a problem in which an agent reads the video turn by turn along the temporal axis, deciding at each turn how fast to watch, what evidence to retain, when to revisit uncertain segments, and when to stop and answer.<n>We propose textbfVideoScout, a multi-turn reasoning agent that instantiates the SEA paradigm through adaptive reasoning pacing.<n>Experiments on long video understanding and reasoning benchmarks demonstrate that our 7B model achieves strong performance compared with existing trained 7B agentic models.
  arXiv  Detail & Related papers  (2026-09-14T14:04:03Z)
• Learning to Drive is a Free Gift: Large-Scale Label-Free Autonomy Pretraining from Unposed In-The-Wild Videos [20.73513310337503]
  Ego-centric driving videos available online provide an abundant source of visual data for autonomous driving.<n>We propose a label-free, teacher-guided framework for learning autonomous driving representations directly from unposed videos.
  arXiv  Detail & Related papers  (2026-02-25T16:38:53Z)
• Repurposing Video Diffusion Transformers for Robust Point Tracking [35.486648006768256]
  Existing methods rely on shallow convolutional backbones such as ResNet that process frames independently.<n>We find that video Transformers (DiTs) inherently exhibit strong point tracking capability and robustly handle dynamic motions.<n>Our work validates video DiT features as an effective and efficient foundation for point tracking.
  arXiv  Detail & Related papers  (2025-12-23T18:54:10Z)
• Dual Learning with Dynamic Knowledge Distillation and Soft Alignment for Partially Relevant Video Retrieval [53.54695034420311]
  In practice, videos are typically untrimmed in long durations with much more complicated background content.<n>We propose a novel framework that distills generalization knowledge from a powerful large-scale vision-language pre-trained model.<n>Experiment results demonstrate that our proposed model achieves state-of-the-art performance on TVR, ActivityNet, and Charades-STA datasets.
  arXiv  Detail & Related papers  (2025-10-14T08:38:20Z)
• ViaRL: Adaptive Temporal Grounding via Visual Iterated Amplification Reinforcement Learning [68.76048244253582]
  We introduce ViaRL, the first framework to leverage rule-based reinforcement learning (RL) for optimizing frame selection in video understanding.<n>ViaRL utilizes the answer accuracy of a downstream model as a reward signal to train a frame selector through trial-and-error.<n>ViaRL consistently delivers superior temporal grounding performance and robust generalization across diverse video understanding tasks.
  arXiv  Detail & Related papers  (2025-05-21T12:29:40Z)
• Seeing Beyond Views: Multi-View Driving Scene Video Generation with Holistic Attention [61.3281618482513]
  We present CogDriving, a novel network designed for synthesizing high-quality multi-view driving videos.<n>CogDriving leverages a Diffusion Transformer architecture with holistic-4D attention modules, enabling simultaneous associations across the dimensions.<n>CogDriving demonstrates strong performance on the nuScenes validation set, achieving an FVD score of 37.8, highlighting its ability to generate realistic driving videos.
  arXiv  Detail & Related papers  (2024-12-04T18:02:49Z)
• RSPNet: Relative Speed Perception for Unsupervised Video Representation Learning [100.76672109782815]
  We study unsupervised video representation learning that seeks to learn both motion and appearance features from unlabeled video only. It is difficult to construct a suitable self-supervised task to well model both motion and appearance features. We propose a new way to perceive the playback speed and exploit the relative speed between two video clips as labels.
  arXiv  Detail & Related papers  (2020-10-27T16:42:50Z)
• AutoTrajectory: Label-free Trajectory Extraction and Prediction from Videos using Dynamic Points [92.91569287889203]
  We present a novel, label-free algorithm, AutoTrajectory, for trajectory extraction and prediction. To better capture the moving objects in videos, we introduce dynamic points. We aggregate dynamic points to instance points, which stand for moving objects such as pedestrians in videos.
  arXiv  Detail & Related papers  (2020-07-11T08:43:34Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.