Question about the WC2014 camera pan,tilt,zoom parameters
gothera opened this issue · comments
So the paper mentions that the images from WC2014 were pre-processed to get the range of these parameters in them:
pan, tilt and focal length ranges
([−35◦,35◦], [−15◦, −5◦] and [1000, 6000] pixels.
Does anyone know more info on how those were got/obtained ?
Also, in synthetic.utils.py, there is the following line:
camera_data = np.asarray([640, 360, 3081.976880,
1.746393, -0.321347, 0.266827,
52.816224, -54.753716, 19.960425])
which I think it's exactly those hard-coded parameters out of nowhere. It would be really helpful to know more info/the approach used to get these for the images in the dataset.
Thanks!
cc @lood339
These parameters are estimated from the training dataset. They are used as the prior (known knowledge).
These parameters are estimated from the training dataset. They are used as the prior (known knowledge).
Hi, could you tell me how to calculate these parameters from dataset ? or share some advance to me. I’m glad to see your reply.