Disclaimer
When you use this material, please cite this paper:
📝 A. Vacavant, T. Chateau, A. Wilhelm, L. Lequievre: A benchmark dataset for outdoor foreground/background extraction. BMC/ACCV 2012.
You may also read and cite those articles:
📝 A. Sobral, A. Vacavant: A comprehensive review of background subtraction algorithms evaluated with synthetic and real videos. CVIU, 122:4–21, 2014.
📝 A. Vacavant, L. Tougne, T. Chateau, L. Robinault: Evaluation of background models with synthetic and real data. In Background Modeling and Foreground Detection for Video Surveillance, 2014.
Learning videos
Those videos have been rendered by Sivic software, with complete ground truth, to test your algorithm and set the best parameters:
🎬 111 - 🎬 111_gt /
🎬 112 - 🎬 112_gt /
🎬 121 - 🎬 122_gt /
🎬 122 - 🎬 122_gt /
🎬 211 - 🎬 211_gt /
🎬 212 - 🎬 212_gt /
🎬 221 - 🎬 221_gt /
🎬 222 - 🎬 222_gt /
🎬 311 - 🎬 311_gt /
🎬 312 - 🎬 312_gt
A complete archive with all videos is available here:
🗃️ bmc_synth1.zip
Evaluation videos
More synthetic videos are provided here, with complete ground truth:
🎬321 - 🎬321_gt /
🎬 322 - 🎬 322_gt /
🎬 411 - 🎬 411_gt /
🎬 412 - 🎬 412_gt /
🎬 421 - 🎬 421_gt /
🎬 422 - 🎬 422_gt /
🎬 511 - 🎬 511_gt /
🎬 512 - 🎬 512_gt /
🎬 521 - 🎬 521_gt /
🎬 522 - 🎬 522_gt
You will also find real long videos in hard contexts, with partial annotations only (see img folder in archives):
🎬 001 -
🎬 002 -
🎬 003 -
🎬 004 -
🎬 005 -
🎬 006 -
🎬 007 -
🎬 008 -
🎬 009
Complete archives with all videos are available there:
🗃️ bmc_synth2.zip - 🗃️ bmc_real.zip