focal_loss.py

import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
from torch.autograd import Variable

class FocalLoss(nn.Module):
    def __init__(self, gamma=0, alpha=None, size_average=True):
        super(FocalLoss, self).__init__()
        self.gamma = gamma
        self.alpha = alpha
        if isinstance(alpha,(float,int)): self.alpha = torch.Tensor([alpha,1-alpha])
        if isinstance(alpha,list): self.alpha = torch.Tensor(alpha)
        self.size_average = size_average

    def forward(self, input, target):
        if input.dim()>2:
            input = input.view(input.size(0),input.size(1),-1)  # N,C,H,W => N,C,H*W
            input = input.transpose(1,2)    # N,C,H*W => N,H*W,C
            input = input.contiguous().view(-1,input.size(2))   # N,H*W,C => N*H*W,C

        # Ignore background pixels

        target_lv = target[:, 0, :, :]
        target_myo = target[:,1,:,:]
        target_rv = target[:,2,:,:]
        target_index = target_myo*1 + target_rv*2
        target_binary_back = target_lv + target_myo + target_rv
        target_binary_back = target_binary_back.view(-1)
        num_objects = torch.sum(target_binary_back)


        target_index = target_index.view(-1,1).long() # N*H*W*C

        logpt = F.log_softmax(input)
        logpt = logpt.gather(1,target_index)
        logpt = logpt.view(-1)
        logpt = logpt * target_binary_back
        pt = Variable(logpt.data.exp())

        if self.alpha is not None:
            if self.alpha.type()!=input.data.type():
                self.alpha = self.alpha.type_as(input.data)
            at = self.alpha.gather(0,target.data.view(-1))
            logpt = logpt * Variable(at)

        loss = -1 * (1-pt)**self.gamma * logpt

        if self.size_average: return loss.sum() / (num_objects.data+1)
        else: return loss.sum()