3D objects Dataset - P.Moreels, 2004
------------------------------------

* This dataset consists of images of objects set on an automated turntable. 
The objects are photographed by 2 cameras set as a stereo system.

* The objects are photographed every 5 degrees, except in the directory 
OneDegreeImageSet/ (every degree) and in the directory Tank/ (every degree).
They are also photographed with 3 different lighting conditions created by 
switching two overhead photographic lights on and off. 
The setup is illustrated in the file Setup3dlab.bmp

* The directories ImageSet070504...ImageSet091504 are similarly organized. 
The reason for having several directories is that the Canon SDK provided to 
control the digital camera was crashing at random times. The camera needed 
then to be turned off and on, and to be recalibrated.

* Each directory contains a Calibration/ subdirectory with images of a checkerboard 
pattern that can be used to calibrate the two cameras. Some directories contain 
Calibration/ and SecondCalibration/ subdirectories. The first one was acquired 
before imaging the objects, the second one was acquired after imaging the objects, 
so that one can check if the conditions were changed between the beginning and 
the end of the image capture.

To calibrate the cameras using these images, you can for example use the matlab calibration
toolbox written by Jean-Yves Bouguet and available at
http://www.vision.caltech.edu/bouguetj/calib_doc/index.html

NOTE: do not pay attention to the .mat and .m files in the calibration directories, they are personal
intermediate calibration results. 

* The subdirectories Bottom/ and Top/ correspond to images from the lower, resp. higher camera.

* File names: images have names of the format img_1-angle_time_#.JPG
where angle characterizes the rotation angle of the turntable 
(0-355 for the first lighting condition, 360-715 for the second lighting condition, 
720-1075 for the third lighting condition).
Time is simply the time of day of the acquisition.
# takes value 0 or 1. Every 90 degrees, the object was imaged twice in the same conditions. 
This was done in order to be able to evaluate the level of noise introduced 
by the camera, jpg compression... these two identical images are indexed by 0 and 1. 
For example, file img_1-540_14-03-52_1.JPG corresponds to an angle of 180 degrees, 
it was taken using the second lighting condition, the time of day the image was taken was 14:03:52, 
and this is the second identical picture taken at this angle with this lighting condition.

File names in the calibration directories - to facilitate the use of the calibration
toolbox mentioned above, the names have been modified to 'calib1.jpg'...'calib66.jpg'.
The original names are still available in file 'oldNames.mat' in each calibration directory.

* The directory 4ScaleImageSet/ contains images of 5 objects taken in similar conditions, 
using 4 different zoom factors on the cameras. 

* Last note: the first few images of the Bannanas are in black/white (wrong command, the bannanas were the first object in the dataset).