Generic torchvision.transform.Compose

vision kornia
1

I wonder if there are any disadvantages of rewriting the default Compose as shown below?

class Compose:
    "Similar to torchvision.transforms.Compose but digest additional kwargs"

    def __init__(self, transforms):
        self.transforms = transforms

    def __call__(self, x, **kwargs):
        for t in self.transforms:
            out = t(x, **kwargs)
            if isinstance(out, Tuple):
                x, kwargs = out
            else:
                x = out
        if isinstance(out, Tuple):
            return x, kwargs
        return x

    def __repr__(self):
        format_string = self.__class__.__name__ + '('
        for t in self.transforms:
            format_string += '\n'
            format_string += '    {0}'.format(t)
        format_string += '\n)'
        return format_string

For something more specialized, review albumination Compose :slight_smile: .