pytorch实现focal loss的两种方式小结
我就废话不多说了，直接上代码吧！
importtorch
importtorch.nn.functionalasF
importnumpyasnp
fromtorch.autogradimportVariable
'''
pytorch实现focalloss的两种方式(现在讨论的是基于分割任务)
在计算损失函数的过程中考虑到类别不平衡的问题，假设加上背景类别共有6个类别
'''
defcompute_class_weights(histogram):
classWeights=np.ones(6,dtype=np.float32)
normHist=histogram/np.sum(histogram)
foriinrange(6):
classWeights[i]=1/(np.log(1.10+normHist[i]))
returnclassWeights
deffocal_loss_my(input,target):
'''
:paraminput:shape[batch_size,num_classes,H,W]仅仅经过卷积操作后的输出，并没有经过任何激活函数的作用
:paramtarget:shape[batch_size,H,W]
:return:
'''
n,c,h,w=input.size()
target=target.long()
input=input.transpose(1,2).transpose(2,3).contiguous().view(-1,c)
target=target.contiguous().view(-1)
number_0=torch.sum(target==0).item()
number_1=torch.sum(target==1).item()
number_2=torch.sum(target==2).item()
number_3=torch.sum(target==3).item()
number_4=torch.sum(target==4).item()
number_5=torch.sum(target==5).item()
frequency=torch.tensor((number_0,number_1,number_2,number_3,number_4,number_5),dtype=torch.float32)
frequency=frequency.numpy()
classWeights=compute_class_weights(frequency)
'''
根据当前给出的groundtruthlabel计算出每个类别所占据的权重
'''
#weights=torch.from_numpy(classWeights).float().cuda()
weights=torch.from_numpy(classWeights).float()
focal_frequency=F.nll_loss(F.softmax(input,dim=1),target,reduction='none')
'''
上面一篇博文讲过
F.nll_loss(torch.log(F.softmax(inputs,dim=1)，target)的函数功能与F.cross_entropy相同
可见F.nll_loss中实现了对于target的one-hotencoding编码功能，将其编码成与inputshape相同的tensor
然后与前面那一项（即F.nll_loss输入的第一项）进行element-wiseproduction
相当于取出了log(p_gt)即当前样本点被分类为正确类别的概率
现在去掉取log的操作，相当于focal_frequencyshape[num_samples]
即取出groundtruth类别的概率数值，并取了负号
'''
focal_frequency+=1.0#shape[num_samples]1-P（gt_classes）
focal_frequency=torch.pow(focal_frequency,2)#torch.Size([75])
focal_frequency=focal_frequency.repeat(c,1)
'''
进行repeat操作后，focal_frequencyshape[num_classes,num_samples]
'''
focal_frequency=focal_frequency.transpose(1,0)
loss=F.nll_loss(focal_frequency*(torch.log(F.softmax(input,dim=1))),target,weight=None,
reduction='elementwise_mean')
returnloss
deffocal_loss_zhihu(input,target):
'''
:paraminput:使用知乎上面大神给出的方案https://zhuanlan.zhihu.com/p/28527749
:paramtarget:
:return:
'''
n,c,h,w=input.size()
target=target.long()
inputs=input.transpose(1,2).transpose(2,3).contiguous().view(-1,c)
target=target.contiguous().view(-1)
N=inputs.size(0)
C=inputs.size(1)
number_0=torch.sum(target==0).item()
number_1=torch.sum(target==1).item()
number_2=torch.sum(target==2).item()
number_3=torch.sum(target==3).item()
number_4=torch.sum(target==4).item()
number_5=torch.sum(target==5).item()
frequency=torch.tensor((number_0,number_1,number_2,number_3,number_4,number_5),dtype=torch.float32)
frequency=frequency.numpy()
classWeights=compute_class_weights(frequency)
weights=torch.from_numpy(classWeights).float()
weights=weights[target.view(-1)]#这行代码非常重要
gamma=2
P=F.softmax(inputs,dim=1)#shape[num_samples,num_classes]
class_mask=inputs.data.new(N,C).fill_(0)
class_mask=Variable(class_mask)
ids=target.view(-1,1)
class_mask.scatter_(1,ids.data,1.)#shape[num_samples,num_classes]one-hotencoding
probs=(P*class_mask).sum(1).view(-1,1)#shape[num_samples,]
log_p=probs.log()
print('incalculatingbatch_loss',weights.shape,probs.shape,log_p.shape)
#batch_loss=-weights*(torch.pow((1-probs),gamma))*log_p
batch_loss=-(torch.pow((1-probs),gamma))*log_p
print(batch_loss.shape)
loss=batch_loss.mean()
returnloss
if__name__=='__main__':
pred=torch.rand((2,6,5,5))
y=torch.from_numpy(np.random.randint(0,6,(2,5,5)))
loss1=focal_loss_my(pred,y)
loss2=focal_loss_zhihu(pred,y)
print('loss1',loss1)
print('loss2',loss2)
'''
incalculatingbatch_losstorch.Size([50])torch.Size([50,1])torch.Size([50,1])
torch.Size([50,1])
loss1tensor(1.3166)
loss2tensor(1.3166)
'''
以上这篇pytorch实现focalloss的两种方式小结就是小编分享给大家的全部内容了，希望能给大家一个参考，也希望大家多多支持毛票票。
声明：本文内容来源于网络，版权归原作者所有，内容由互联网用户自发贡献自行上传，本网站不拥有所有权，未作人工编辑处理，也不承担相关法律责任。如果您发现有涉嫌版权的内容，欢迎发送邮件至：czq8825#qq.com（发邮件时，请将#更换为@）进行举报，并提供相关证据，一经查实，本站将立刻删除涉嫌侵权内容。