def loss_labels(self, outputs, targets, indices, num_boxes, log=True):
"""Classification loss (NLL)
targets dicts must contain the key "labels" containing a tensor of dim [nb_target_boxes]
"""
assert "pred_logits" in outputs
del num_boxes
src_logits = outputs["pred_logits"]
idx = self._get_src_permutation_idx(indices)
target_classes_o = torch.cat(
[t["labels"][J] for t, (_, J) in zip(targets, indices)])
target_classes = torch.full(src_logits.shape[:2],
self.num_classes,
dtype=torch.int64,
device=src_logits.device)
target_classes[idx] = target_classes_o
loss_ce = F.cross_entropy(src_logits.transpose(1, 2), target_classes,
self.empty_weight)
losses = {"loss_ce": loss_ce}
if log:
# TODO this should probably be a separate loss, not hacked in this one here
losses["class_error"] = 100 - accuracy(src_logits[idx],
target_classes_o)[0]
return losses
def loss_labels(self, outputs, targets, indices, num_boxes, log=True):
"""Classification loss (NLL)
targets dicts must contain the key "labels" containing a tensor of dim [nb_target_boxes]
"""
assert "pred_logits" in outputs
src_logits = outputs["pred_logits"]
idx = self._get_src_permutation_idx(indices)
target_classes_o = torch.cat([t["labels"][J] for t, (_, J) in zip(targets, indices)])
target_classes = torch.full(
src_logits.shape[:2], self.num_classes, dtype=torch.int64, device=src_logits.device
)
target_classes[idx] = target_classes_o
if self.use_focal:
flat_src_logits = src_logits.flatten(0, 1)
target_classes = target_classes.flatten(0, 1)
pos_inds = torch.nonzero(target_classes != self.num_classes, as_tuple=True)[0]
labels = torch.zeros_like(flat_src_logits)
labels[pos_inds, target_classes[pos_inds]] = 1
# comp focal loss.
class_loss = sigmoid_focal_loss_jit(
flat_src_logits,
labels,
alpha=self.focal_loss_alpha,
gamma=self.focal_loss_gamma,
reduction="sum",
) / num_boxes
losses = {'loss_ce': class_loss}
else:
loss_ce = F.cross_entropy(src_logits.transpose(1, 2), target_classes, self.empty_weight)
losses = {"loss_ce": loss_ce}
if log:
losses["class_error"] = 100 - accuracy(src_logits[idx], target_classes_o)[0]
return losses