Question regarding color constancy loss

[bsun0802]

Hi Author,

In the MyLoss.py you uploaded,

class L_color(nn.Module):

    def __init__(self):
        super(L_color, self).__init__()

    def forward(self, x ):

        b,c,h,w = x.shape

        mean_rgb = torch.mean(x,[2,3],keepdim=True)
        mr,mg, mb = torch.split(mean_rgb, 1, dim=1)
        Drg = torch.pow(mr-mg,2)
        Drb = torch.pow(mr-mb,2)
        Dgb = torch.pow(mb-mg,2)
        k = torch.pow(torch.pow(Drg,2) + torch.pow(Drb,2) + torch.pow(Dgb,2),0.5)

        return k

The Drg was powered to 2 when Drg was defined, and when returning k, the Drg was powered to 2 again, so the difference is indeed powered to 4. Is it intentional design? and what is the reason for that?

[Li-Chongyi]

Hi,

There is no natural difference. You can just use ^2, but you have to change the weight of color loss in the training phase.

BEST,

Bo Sun notifications@github.com 于2020年7月25日周六 下午10:09写道：

Hi Author,

In the MyLoss.py you uploaded,

class L_color(nn.Module):

def __init__(self):
    super(L_color, self).__init__()

def forward(self, x ):

    b,c,h,w = x.shape

    mean_rgb = torch.mean(x,[2,3],keepdim=True)
    mr,mg, mb = torch.split(mean_rgb, 1, dim=1)
    Drg = torch.pow(mr-mg,2)
    Drb = torch.pow(mr-mb,2)
    Dgb = torch.pow(mb-mg,2)
    k = torch.pow(torch.pow(Drg,2) + torch.pow(Drb,2) + torch.pow(Dgb,2),0.5)

    return k

The Drg was powered to 2 when Drg was defined, and when returning k, the Drg was powered to 2 again, so the difference is indeed powered to 4. Is it intentional design? and what is the reason for that?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/Li-Chongyi/Zero-DCE/issues/6, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIEVQEOFG76WLLMXBP7TRCDR5LRR3ANCNFSM4PHPG3SQ .