Image-Adaptive-3DLUT
Image-Adaptive-3DLUT copied to clipboard
How does it work with "Generator3DLUT_identity"
In the model.py and the code for training,
What is the difference of the "Generator3DLUT_identity" and "Generator3DLUT_zero" ?
At first, in my opinion, the parameters of the N=3 LUTs are learned while training. However, why is the "Generator3DLUT_identity" be designed? Looking forward to some reply!
It's a good job actually!
In the model.py and the code for training,
What is the difference of the "Generator3DLUT_identity" and "Generator3DLUT_zero" ?
At first, in my opinion, the parameters of the N=3 LUTs are learned while training. However, why is the "Generator3DLUT_identity" be designed? Looking forward to some reply!
It's a good job actually!
As the paper mentioned in section 3.5, “such an initialization ensures that the initial output of our model is not far away from the input and thus increases the training speed”. (* ̄︶ ̄)
Hi, the identity LUT itself outputs the same image as the input while the zero LUT output zeros for all RGB inputs.