=== BOATS DATASET ===

Antoni Chan (abchan at cityu dot edu dot hk)
Dept. Computer Science, City University of Hong Kong.
Statistical Visual Computing Lab, University of California, San Diego.


-- Notes --
These are the boats videos used in [1] for testing background subtraction
in water scenes.  There are 3 videos: 
  1) Boats0 is the training video with just the water background.
  2) Boats1 contains two boats in the distance.
  3) Boats2 contains one boat that is close to the camera.
These are consecutive video clips from the same scene.  Three formats are 
provided: 
  1) RGB color with a frame size of 360x200.
  2) grayscale with a frame size of 360x200.
  3) grayscale with a frame size of 180x100.
All videos are 15 fps, and saved as a sequence of png files.  
The smallest format (180x100 grayscale) is used in [1].

Ground-truth foreground masks are also provided for the smaller videos.
The masks contain three regions:
  0   = background,
  128 = mixed foreground and background, 
  255 = foreground.  
In [1], the quantitative results ignore the mixed type regions, since 
the true foreground/background state is ambignuous, and too tedious
to mark by hand.


-- Videos --

Each video is saved as a sequence of png files.  The "fXXX" in the 
filename indicates the frame number.  
The following directories contain the frames for each video:

Boats0_rgb/           -- Boats0, color, 360x200, 200 frames
Boats1_rgb/           -- Boats1, color, 360x200, 300 frames
Boats2_rgb/           -- Boats2, color, 360x200, 300 frames

Boats0_y/             -- Boats0, gray, 360x200, 200 frames
Boats1_y/             -- Boats1, gray, 360x200, 300 frames
Boats2_y/             -- Boats2, gray, 360x200, 300 frames

Boats0_small_y/       -- Boats0, gray, 180x100, 200 frames
Boats1_small_y/       -- Boats1, gray, 180x100, 300 frames
Boats2_small_y/       -- Boats2, gray, 180x100, 300 frames

Boats1_small_fgmask/  -- Boats1, foreground mask, 180x100, 300 frames
Boats2_small_fgmask/  -- Boats2, foreground mask, 180x100, 300 frames


-- References --
[1] A.B. Chan, V. Mahadevan, and N. Vasconcelos.
    "Generalized Stauffer-Grimson Background Subtraction for Dynamic Scenes",
    Machine Vision and Applications, vol 22(5), pp 751-766, Sept 2011.


-- Version --
2013-01-07: version 3 - use png files, readme updated
2013-01-04: version 2 - readme updated
2010-11-17: version 1


