Two-Pass Zero-Shot Temporal-Spatial Grounding of Rare Traffic Events in Surveillance Video
Not on the map. Why?
Abstract (arXiv)
Grounding traffic accidents in real CCTV footage is a rare-event problem where training on labeled accident video is often prohibited, yet accurate joint localization in time, space, and collision type is required. We present a no-fine-tuning pipeline that elicits this joint output from frozen vision-language models through two ideas. First, a coarse-to-fine two-pass decomposition: a full-video pass at 1 fps produces a coarse (t, x, y, c) tuple, then a second pass at 5 fps within a +/- 3 s window refines time and location, with two deterministic confidence gates that revert to the coarse estimate on boundary hedges or edge-clamped coordinates. Second, a specialist role assignment: Qwen3-VL-Plus handles grounding, Gemini 3.1 Flash-Lite handles typing on a centered video clip. On the ACCIDENT@CVPR 2026 benchmark (2,027 real CCTV videos) we reach ACCS = 0.539 (95% CI [0.525, 0.553]): +0.127 over the benchmark paper's best-of-baselines oracle (0.412), +0.143 over the strongest single-VLM baseline (Molmo-7B, 0.396), and +0.250 over the naive baseline (0.289). The VLM path uses up to three API calls per video (17% fall back to physics on API failures); the full run costs ~20.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 18 of its 27 references that are in the map. First posted on arXiv 2 May 2026.
Builds on
18 of 27 references.
References · 18
- 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
- 2025Qwen3-VL Technical Reportcited by 2,793
- 2025SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Featurescited by 1,188
- 2024Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Modelscited by 329
- 2023VTimeLLM: Empower LLM to Grasp Video Momentscited by 255
- 2024VideoTree: Adaptive Tree-based Video Representation for LLM Reasoning on Long Videoscited by 251
- 2023VadCLIP: Adapting Vision-Language Models for Weakly Supervised Video Anomaly Detectioncited by 123
- 2020Uncertainty-based Traffic Accident Anticipation with Spatio-Temporal Relational Learningcited by 91
- 2024Grounded-VideoLLM: Sharpening Fine-grained Temporal Grounding in Video Large Language Modelscited by 91
- 2024TRACE: Temporal Grounding Video LLM via Causal Event Modelingcited by 89
- 2024Harnessing Large Language Models for Training-free Video Anomaly Detectioncited by 87
- 2021A Dynamic Spatial-temporal Attention Network for Early Anticipation of Traffic Accidentscited by 42
6 more
- 2018CADP: A Novel Dataset for CCTV Traffic Camera based Accident Analysiscited by 39
- 2023Vision-Based Traffic Accident Detection and Anticipation: A Surveycited by 31
- 2024ReVisionLLM: Recursive Vision-Language Model for Temporal Grounding in Hour-Long Videoscited by 18
- 2025VideoMiner: Iteratively Grounding Key Frames of Hour-Long Videos via Tree-based Group Relative Policy Optimizationcited by 11
- 2026ACCIDENT: A Benchmark Dataset for Vehicle Accident Detection from Traffic Surveillance Videoscited by 7
- 2026A Modular Zero-Shot Pipeline for Accident Detection, Localization, and Classification in Traffic Surveillance Videocited by 2
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026Zero-Shot Traffic Accident Detection via a Coarse-to-Fine VLM-Tracking Pipeline
- 2026Metadata-Aware Multi-Prompt Reasoning for Zero-Shot Accident Understanding
- 2025Enhancing Traffic Incident Response through Sub-Second Temporal Localization with HybridMamba
- 2024Training-free Video Temporal Grounding using Large-scale Pre-trained Models
- 2026Efficient Spatio-Temporal Grounding with Multimodal Large Models via Second-Level Tracking and RL Verification
- 2026From Detection to Understanding: TAR and TAR-Bench for Multi-Task Traffic Anomaly Reasoning
- 2026From Accuracy to Visual Dependence: Auditing and Filtering Modality Collapse in Traffic VideoQA
- 2025TUMTraffic-VideoQA: A Benchmark for Unified Spatio-Temporal Video Understanding in Traffic Scenes