CrossEntropyLoss2d

[kshpv]

Hello!
Thank you for your work!
I am a little bit confused about CrossEntropyLoss2d class, especially:

targets_m[mask] -= 1
loss_all = self.ce_loss(inputs, targets_m.long())
losses.append(torch.sum(torch.masked_select(loss_all, mask)) / torch.sum(mask.float()))

1. Why do you subtract 1 from targets masks?
2. Why do you use torch.masked_select(loss_all, mask)) for calculating a loss for a single target? It seems incorrect because in that case we forget about a background.

Thank you!

[kshpv]

By the way, Why do you add 1 to all class labels in grid_image = make_grid(utils.color_label(torch.max(pred_scales[0][:3], 1)[1] + 1), 3, normalize=False, range=(0, 255))?
Is it correct?

Thank you!