监控视频中罕见交通事件的两次通过零样本时空定位
Two-Pass Zero-Shot Temporal-Spatial Grounding of Rare Traffic Events in Surveillance Video
AI总结：
提出一种无需微调的管道，通过粗到细的两遍分解和专家角色分配，利用冻结视觉语言模型实现罕见交通事件在时间、空间和碰撞类型上的联合定位。
摘要：
在真实闭路电视画面中定位交通事故是一个罕见事件问题，通常禁止使用标注事故视频进行训练，但需要精确的时空和碰撞类型联合定位。我们提出一种无需微调的管道，通过两个想法从冻结的视觉语言模型中引出这种联合输出。首先，粗到细的两遍分解：第一遍以1 fps处理全视频，产生粗粒度(t, x, y, c)元组；然后第二遍在±3秒窗口内以5 fps细化时间和位置，并设置两个确定性置信门，在边界犹豫或边缘夹紧坐标时回退到粗估计。其次，专家角色分配：Qwen3-VL-Plus负责定位，Gemini 3.1 Flash-Lite负责在居中视频片段上分类。在ACCIDENT@CVPR 2026基准测试（2,027个真实闭路电视视频）上，我们达到ACC^S = 0.539（95%置信区间[0.525, 0.553]）：比基准论文的最佳基线预言机（0.412）高0.127，比最强单VLM基线（Molmo-7B, 0.396）高0.143，比朴素基线（0.289）高0.250。VLM路径每个视频最多调用三次API（17%在API失败时回退到物理方法）；完整运行成本约20美元。
英文摘要：
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 ACC^S = 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.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
