UCSD Pedestrian Database
Statistical Visual Computing Lab

ucsdpeds.zip791.87MB
Type: Dataset
Tags:

Bibtex:
@article{,
title= {UCSD Pedestrian Database},
journal= {},
author= {Statistical Visual Computing Lab},
year= {2008},
url= {http://www.svcl.ucsd.edu/projects/peoplecnt/db/readme.pdf},
abstract= {This is the UCSD pedestrian database used in “Modeling, Clustering, and Segmenting Video with Mixtures of Dynamic Textures”

## Database Format

The database contains video of pedestrians on UCSD walkways, taken from a stationary camera.
All videos are 8-bit grayscale, with dimensions 238 × 158 at 10 fps. The database is split into
scenes, taken from different viewpoints (currently, only one scene is available...more are coming).
Each scene is in its own directory vidX where X is a letter (e.g. vidf), and is split into video clips
of length 200 named vidfXY 33 ZZZ.y, where Y and ZZZ are numbers. Finally, each video clip is
saved as a set of .png files. Examples from each scene are presented in Figure 1. If you use this
database, please reference.


![](https://i.imgur.com/VTk15Rx.png)},
keywords= {},
terms= {}
}

Report