A combination of background modeler and encoder-decoder CNN for background/foreground segregation in image sequence