Description

PETS (Performance Evaluation of Tracking and Surveillance) is a long-running series of public surveillance video benchmarks released yearly from the early 2000s onward. The most-cited editions in the vault are PETS 2001 (object tracking), PETS 2006 / 2007 (left-luggage and event detection), and PETS 2009 (multi-camera crowd flow). It is the standard benchmark family for trajectory-based event detection, abnormal-event detection, and surveillance-scenario querying.

Modality / size

  • Modality: multi-camera surveillance video.
  • Subjects / scenarios: train stations, public spaces; varies per edition.
  • Labels: event / abnormality annotations and pedestrian trajectories.

Used by (papers)

  • Tracking and abnormal-event detection baselines; cited next to UMN, MIT traffic, and similar surveillance datasets.

Notes

  • Public datasets; available from the PETS workshop archives.
  • Multiple editions are folded into this note since papers cite them interchangeably.

1 vault paper evaluate on this dataset

Titles and DOIs only — no abstracts, no analyses.

  • Recent trends in crowd analysis: A review 2021 DOI ↗