在 unet 架构中使用自定义权重图的正确方法

Question

u-net 架构中有一个著名的技巧，就是使用自定义权重图来提高准确性。下面是它的详细信息：

现在，通过在这里和其他多个地方询问，我了解了 2 种方法。我想知道哪个是正确的，或者还有其他更正确的正确方法吗？

1. 首先是torch.nn.Functional在训练循环中使用方法：

   loss = torch.nn.functional.cross_entropy(output, target, w) 其中 w 将是计算出的自定义重量。

2. 二是reduction='none'在训练循环外调用损失函数时 使用criterion = torch.nn.CrossEntropy(reduction='none')

   然后在训练循环中乘以自定义权重：

   gt # Ground truth, format torch.long
   pd # Network output
   W # per-element weighting based on the distance map from UNet
   loss = criterion(pd, gt)
   loss = W*loss # Ensure that weights are scaled appropriately
   loss = torch.sum(loss.flatten(start_dim=1), axis=0) # Sums the loss per image
   loss = torch.mean(loss) # Average across a batch
   

现在，我有点困惑哪个是正确的，或者还有其他方法，还是两者都是正确的？

Answer 1

加权部分看起来只是简单的加权交叉熵，它是针对类的数量（在下面的示例中为 2）执行的。

weights = torch.FloatTensor([.3, .7])
loss_func = nn.CrossEntropyLoss(weight=weights)

编辑：

您看过Patrick Black 的这个实现吗？

# Set properties
batch_size = 10
out_channels = 2
W = 10
H = 10

# Initialize logits etc. with random
logits = torch.FloatTensor(batch_size, out_channels, H, W).normal_()
target = torch.LongTensor(batch_size, H, W).random_(0, out_channels)
weights = torch.FloatTensor(batch_size, 1, H, W).random_(1, 3)

# Calculate log probabilities
logp = F.log_softmax(logits)

# Gather log probabilities with respect to target
logp = logp.gather(1, target.view(batch_size, 1, H, W))

# Multiply with weights
weighted_logp = (logp * weights).view(batch_size, -1)

# Rescale so that loss is in approx. same interval
weighted_loss = weighted_logp.sum(1) / weights.view(batch_size, -1).sum(1)

# Average over mini-batch
weighted_loss = -1. * weighted_loss.mean()