分析这个代码class OhemCrossEntropy(nn.Module): def __init__(self, ignore_label=-1, thres=0.7, min_kept=100000, weight=None): super(OhemCrossEntropy, self).__init__() self.thresh = thres self.min_kept = max(1, min_kept) self.ignore_label = ignore_label self.criterion = nn.CrossEntropyLoss( weight=weight, ignore_index=ignore_label, reduction='none' ) def _ce_forward(self, score, target): ph, pw = score.size(2), score.size(3) h, w = target.size(1), target.size(2) if ph != h or pw != w: score = F.interpolate(input=score, size=( h, w), mode='bilinear', align_corners=config.MODEL.ALIGN_CORNERS) loss = self.criterion(score, target) return loss def _ohem_forward(self, score, target, **kwargs): ph, pw = score.size(2), score.size(3) h, w = target.size(1), target.size(2) if ph != h or pw != w: score = F.interpolate(input=score, size=( h, w), mode='bilinear', align_corners=config.MODEL.ALIGN_CORNERS) pred = F.softmax(score, dim=1) pixel_losses = self.criterion(score, target).contiguous().view(-1) mask = target.contiguous().view(-1) != self.ignore_label tmp_target = target.clone() tmp_target[tmp_target == self.ignore_label] = 0 pred = pred.gather(1, tmp_target.unsqueeze(1)) pred, ind = pred.contiguous().view(-1,)[mask].contiguous().sort() min_value = pred[min(self.min_kept, pred.numel() - 1)] threshold = max(min_value, self.thresh) pixel_losses = pixel_losses[mask][ind] pixel_losses = pixel_losses[pred < threshold] return pixel_losses.mean() def forward(self, score, target): if config.MODEL.NUM_OUTPUTS == 1: score = [score] weights = config.LOSS.BALANCE_WEIGHTS assert len(weights) == len(score) functions = [self._ce_forward] * \ (len(weights) - 1) + [self._ohem_forward] return sum([ w * func(x, target) for (w, x, func) in zip(weights, score, functions) ])
时间: 2023-06-10 14:03:24 浏览: 219
查看class代码
这是一个实现了OHEM(Online Hard Example Mining)的交叉熵损失函数,用于解决深度学习中存在难样本或噪声样本导致训练效果不好的问题。其中,thresh表示像素的softmax预测概率阈值,小于该阈值的像素被认为是难样本;min_kept表示每张图中保留的难样本最小数量;ignore_label表示忽略的标签;weight表示权重。_ce_forward函数实现了普通的交叉熵损失的计算;_ohem_forward函数实现了OHEM的计算。在forward函数中,对于有多个输出的模型,采用了权重平衡的方式进行计算。
阅读全文