Some questions about the ‘bce_iou_loss’ function

[Liqq1]

def bce_iou_loss(pred, mask):
    weight = 1 + 5 * torch.abs(F.avg_pool2d(mask, kernel_size=31, stride=1, padding=15) - mask)

    bce = F.binary_cross_entropy_with_logits(pred, mask, reduction='none')

    pred = torch.sigmoid(pred)
    inter = pred * mask
    union = pred + mask
    iou = 1 - (inter + 1) / (union - inter + 1)

    weighted_bce = (weight * bce).sum(dim=(2, 3)) / weight.sum(dim=(2, 3))
    weighted_iou = (weight * iou).sum(dim=(2, 3)) / weight.sum(dim=(2, 3))

    return (weighted_bce + weighted_iou).mean()

I am looking forward to your reply！

[plemeri]

Sorry for the late reply. As we used PraNet as our baseline, we thought the loss function was also from them. In Fact, the authors of PraNet referred F3Net which proposed the loss function that we used, and we should have referred them too.

Long story short, sorry that we mislead some information, please refer to F3Net for more information.

[Liqq1]

Thanks for your reply, I will refer to F3Net for more information～