Penn-Fudan Database for Pedestrian Detection and Segmentation

这是一个图像数据库，其中包含在 [1] 中报告的实验中用于行人检测的图像。这些图像取自校园和城市街道周围的场景。我们对这些图像感兴趣的对象是行人。每个图像中至少有一个行人。该数据库中标记行人的高度属于 [180,390] 像素。所有带标签的行人都是笔直向上的，有345个带标签的行人的图像有170个，其中96个图像取自宾夕法尼亚大学周围，其他74个图像取自复旦大学周围。

This is an image database containing the images employed for pedestrian detection in the experiments reported in [1]. These images were acquired from scenes surrounding university campuses and urban streets. The target objects of interest in these images are pedestrians, and every image includes at least one pedestrian. The height of all labeled pedestrians in this database ranges from 180 to 390 pixels. All labeled pedestrians are in an upright orientation. There are 170 images in total, which collectively contain 345 labeled pedestrian instances. Among them, 96 images were collected around the University of Pennsylvania, while the remaining 74 images were captured around Fudan University.

搜集汇总

背景与挑战

背景概述

该数据集是一个用于行人检测和分割的图像数据库，包含来自宾夕法尼亚大学和复旦大学周围的场景图像，所有图像中至少有一个行人，行人高度在180到390像素之间。数据集由两所大学于2007年联合发布。