Pedestrian Detection Database

Abstraction: Penn-Fudan annotated pedestrian detection benchmark image dataset

Key points:

• Dataset contains 170 images with 345 labeled pedestrians; 96 images from University of Pennsylvania campus, 74 from Fudan University surroundings.
• Labeled pedestrian heights range from 180 to 390 pixels; all labeled pedestrians are upright.
• Annotation format is compatible with PASCAL VOC Annotation Version 1.00, enabling broad toolchain compatibility.
• Download is ~51 MB zipped (~52 MB unzipped); hosted at CIS, University of Pennsylvania.
• Used in published pedestrian detection experiments; each image contains at least one pedestrian.

Connections: University Of Pennsylvania · Fudan University · Pedestrian Detection · Object Detection · Computer Vision

Source: https://www.cis.upenn.edu/~jshi/ped_html/