Welcome to The Longterm Observation of Scenes with Tracks Dataset

The LOST Dataset comprises videos taken from streaming outdoor webcams, capturing the same half hour each day. Data collection started in June of 2010 and continues to this day. LOST contains rich metadata, including geolocation, object detections, and tracking results.

To date, we have captured 17,100 videos, for a total of 155,029,134 frames and 727 Gb of compressed video data. Watching it all would take over 258 days, 2 hours, and 24 minutes. Processing this data has given us 434,046,047 object detections and 1,243,475 tracks.

We believe that sharing this dataset opens opportunities for computer vision research involving very long-term outdoor surveillance, robust anomaly detection, and scene analysis methods based on trajectories.

Acknowledgements

This project is supported under

  • NSF IIS 0546383: "CAREER: Passive Vision, What Can Be Learned by a Stationary Observer"
  • DARPA Grant D11AP00255: "ContextualEyes: A Context-Aware Surveillance System"
Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation or DARPA.