Appendix

Channel Names

The following 19 Nuke channels can be used when setting the Channels In and Channels Out knobs in the CatFileCreator node:

• rgba.red

• rgba.green

• rgba.blue

• rgba.alpha

• depth.Z

• forward.u

• forward.v

• backward.u

• backward.v

• deep.front

• deep.back

• disparityL.x

• disparityL.y

• disparityR.x

• disparityR.y

• mask.a

• rotopaint_mask.a

• mask_planartrack.a

• mask_splinewarp.a

Supported Nuke Knobs

The following 6 Nuke knobs can be added as a custom knob in the CatFileCreator:

Integer knob (Int_Knob)

Class attributes controlled by an integer knob in Nuke should be defined as integers in your PyTorch model:

class Cifar10ResNet(torch.nn.Module):
   def __init__(self, int_knob = 0):

Floating point slider (Float_Knob)

Class attributes controlled by a float knob in Nuke should be defined as floats in your PyTorch model:

class Cifar10ResNet(torch.nn.Module):
   def __init__(self, float_knob = 1.0):

2D Position knob (XY_Knob)

Class attributes controlled by a 2D position knob in Nuke should be defined as tensors in your PyTorch model:

class Cifar10ResNet(torch.nn.Module):
   def __init__(self, xy_knob = torch.tensor([0, 0])):

3D Position knob (XYZ_Knob)

Class attributes controlled by a 3D position knob in Nuke should be defined as tensors in your PyTorch model:

class Cifar10ResNet(torch.nn.Module):
   def __init__(self,  xyz_knob = torch.tensor([0, 0, 0])):

Check Box (Bool_Knob)

Class attributes controlled by a checkbox in Nuke should be defined as an integer with values 0 or 1 in your PyTorch model:

class Cifar10ResNet(torch.nn.Module):
   def __init__(self, bool_knob = 0):

Pulldown choice (Enumeration_Knob)

Class attributes controlled by a pulldown choice knob in Nuke should be defined as an integer with values between 0 and (num_options - 1), where num_options is the number of options you have defined in your pulldown choice knob in Nuke:

class Cifar10ResNet(torch.nn.Module):
   def __init__(self, enum_knob = 3):