Boston University - VIP Lab: This video shows the output of BSUV-Net 2.0 background subtraction algorithm applied to a night-time video clip recorded by a traffic camera in Tokyo-Shinjuku, Japan (www.youtube.com/watch?v=RQA5RcIZlAM). The original recording is shown on the left, foreground/background prediction is shown in the middle and their combination (darker pixels represent background regions, brighter pixels represent foreground regions) is shown on the right. Note that the flashing billboards are ignored thanks to spatio-temporal data augmentation methods used during training. For details, see [ Ссылка ]

свернуть