医学图像分割损失函数Dice

class DiceLoss(nn.Module):
    def __init__(self):
        super(DiceLoss, self).__init__()
    def forward(self, input, targets):
        # 获取每个批次的大小 N
        N = targets.size()[0]
        # 平滑变量
        smooth = 1
        # 将宽高 reshape 到同一纬度
        input_flat = input.view(N, -1)
        targets_flat = targets.view(N, -1)
        # 计算交集
        intersection = input_flat * targets_flat
        dice_eff = (2 * intersection.sum(1) + smooth) / (input_flat.sum(1) + targets_flat.sum(1) + smooth)
        # 计算一个批次中平均每张图的损失
        dice_loss = 1 - dice_eff.sum() / N

        return dice_loss

医学图像分割最常用的损失函数。