video-subtitle-remover/backend/tools/train/loss_sttn.py

import torch
import torch.nn as nn


class AdversarialLoss(nn.Module):
    r"""
    Adversarial loss
    https://arxiv.org/abs/1711.10337
    """

    def __init__(self, type='nsgan', target_real_label=1.0, target_fake_label=0.0):
        r"""
        type = nsgan | lsgan | hinge
        """
        super(AdversarialLoss, self).__init__()
        self.type = type
        self.register_buffer('real_label', torch.tensor(target_real_label))
        self.register_buffer('fake_label', torch.tensor(target_fake_label))

        if type == 'nsgan':
            self.criterion = nn.BCELoss()
        elif type == 'lsgan':
            self.criterion = nn.MSELoss()
        elif type == 'hinge':
            self.criterion = nn.ReLU()

    def __call__(self, outputs, is_real, is_disc=None):
        if self.type == 'hinge':
            if is_disc:
                if is_real:
                    outputs = -outputs
                return self.criterion(1 + outputs).mean()
            else:
                return (-outputs).mean()
        else:
            labels = (self.real_label if is_real else self.fake_label).expand_as(
                outputs)
            loss = self.criterion(outputs, labels)
            return loss