Caltech Pedestrian Detection Benchmark 步行数据集

Caltech Pedestrian Detection Benchmark Dataset 是一个用于检测行人的数据集，其包含约 10 小时 640 * 480 的 30Hz 视频，主要由行驶在乡村街道上的小车拍摄，视频共计约 250,000 帧，包含 350,000 个边界框和 2300 个行人的注释，其中注释包括包围盒详细的闭塞标签之间的对应关系。

The Caltech Pedestrian Detection Benchmark Dataset is a dedicated dataset for pedestrian detection research. It contains approximately 10 hours of 30Hz video with a resolution of 640 × 480, most of which was captured by cars driving on rural streets. The dataset includes a total of roughly 250,000 video frames, along with 350,000 bounding boxes and annotations for 2,300 pedestrians. The annotations cover detailed occlusion labels and the correspondence between each bounding box and its corresponding occlusion label.

搜集汇总

背景与挑战

背景概述

Caltech Pedestrian Detection Benchmark是一个用于行人检测的经典数据集，包含约10小时30Hz的街道视频(250,000帧)，标注了350,000个边界框和2,300个行人。该数据集由加州理工学院于2009年发布，特别标注了行人遮挡关系，适用于安防和机器视觉研究。