# Two-Pass Zero-Shot Temporal-Spatial Grounding of Rare Traffic Events in Surveillance Video

arXiv: https://arxiv.org/abs/2605.01512 · HTML: https://mlsanity.com/paper/2605.01512/

## 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 (most cited first, 18 of 18)

- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features](https://mlsanity.com/paper/2502.14786/) (2025)
- [Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models](https://mlsanity.com/paper/2409.17146/) (2024)
- [VTimeLLM: Empower LLM to Grasp Video Moments](https://mlsanity.com/paper/2311.18445/) (2023)
- [VideoTree: Adaptive Tree-based Video Representation for LLM Reasoning on Long Videos](https://mlsanity.com/paper/2405.19209/) (2024)
- [VadCLIP: Adapting Vision-Language Models for Weakly Supervised Video Anomaly Detection](https://mlsanity.com/paper/2308.11681/) (2023)
- [Uncertainty-based Traffic Accident Anticipation with Spatio-Temporal Relational Learning](https://mlsanity.com/paper/2008.00334/) (2020)
- [Grounded-VideoLLM: Sharpening Fine-grained Temporal Grounding in Video Large Language Models](https://mlsanity.com/paper/2410.03290/) (2024)
- [TRACE: Temporal Grounding Video LLM via Causal Event Modeling](https://mlsanity.com/paper/2410.05643/) (2024)
- [Harnessing Large Language Models for Training-free Video Anomaly Detection](https://mlsanity.com/paper/2404.01014/) (2024)
- [A Dynamic Spatial-temporal Attention Network for Early Anticipation of Traffic Accidents](https://mlsanity.com/paper/2106.10197/) (2021)
- [CADP: A Novel Dataset for CCTV Traffic Camera based Accident Analysis](https://mlsanity.com/paper/1809.05782/) (2018)
- [Vision-Based Traffic Accident Detection and Anticipation: A Survey](https://mlsanity.com/paper/2308.15985/) (2023)
- [ReVisionLLM: Recursive Vision-Language Model for Temporal Grounding in Hour-Long Videos](https://mlsanity.com/paper/2411.14901/) (2024)
- [VideoMiner: Iteratively Grounding Key Frames of Hour-Long Videos via Tree-based Group Relative Policy Optimization](https://mlsanity.com/paper/2510.06040/) (2025)
- [ACCIDENT: A Benchmark Dataset for Vehicle Accident Detection from Traffic Surveillance Videos](https://mlsanity.com/paper/2604.09819/) (2026)
- [A Modular Zero-Shot Pipeline for Accident Detection, Localization, and Classification in Traffic Surveillance Video](https://mlsanity.com/paper/2604.09685/) (2026)

## 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.