TV_Loss实现
原理介绍
import torch
from torch.autograd import Variable
class TVLoss(torch.nn.Module):
"""
TV loss
"""
def __init__(self, weight=1):
super(TVLoss, self).__init__()
self.weight = weight
def forward(self, x):
batch_size = x.size()[0]
h_x = x.size()[2]
w_x = x.size()[3]
count_h = self._tensor_size(x[:, :, 1:, :])
count_w = self._tensor_size(x[:, :, :, 1:])
h_tv = torch.pow((x[:, :, 1:, :] - x[:, :, :h_x - 1, :]), 2).sum()
w_tv = torch.pow((x[:, :, :, 1:] - x[:, :, :, :w_x - 1]), 2).sum()
return self.weight * 2 * (h_tv / count_h + w_tv / count_w) / batch_size
def _tensor_size(self, t):
return t.size()[1] * t.size()[2] * t.size()[3]
if __name__ == "__main__":
x = Variable(
torch.FloatTensor([[[1, 2, 3], [2, 3, 4], [3, 4, 5]], [[1, 2, 3], [2, 3, 4], [3, 4, 5]]]).view(1, 2, 3, 3),
requires_grad=True)
addition = TVLoss()
z = addition(x)
print(z)
z.backward()