def loss(self, mask_pred, mask_targets, labels):
num_pos = labels.new_ones(labels.size()).float().sum()
avg_factor = torch.clamp(reduce_mean(num_pos), min=1.).item()
loss = dict()
if mask_pred.size(0) == 0:
loss_mask = mask_pred.sum()
else:
loss_mask = self.loss_mask(
mask_pred[torch.arange(num_pos).long(), labels, ...].sigmoid(),
mask_targets,
avg_factor=avg_factor)
loss['loss_mask'] = loss_mask
return loss
def loss(self,
cls_score,
bbox_pred,
labels,
label_weights,
bbox_targets,
bbox_weights,
imgs_whwh=None,
reduction_override=None,
**kwargs):
""""Loss function of DIIHead, get loss of all images.
Args:
cls_score (Tensor): Classification prediction
results of all class, has shape
(batch_size * num_proposals_single_image, num_classes)
bbox_pred (Tensor): Regression prediction results,
has shape
(batch_size * num_proposals_single_image, 4), the last
dimension 4 represents [tl_x, tl_y, br_x, br_y].
labels (Tensor): Label of each proposals, has shape
(batch_size * num_proposals_single_image
label_weights (Tensor): Classification loss
weight of each proposals, has shape
(batch_size * num_proposals_single_image
bbox_targets (Tensor): Regression targets of each
proposals, has shape
(batch_size * num_proposals_single_image, 4),
the last dimension 4 represents
[tl_x, tl_y, br_x, br_y].
bbox_weights (Tensor): Regression loss weight of each
proposals's coordinate, has shape
(batch_size * num_proposals_single_image, 4),
imgs_whwh (Tensor): imgs_whwh (Tensor): Tensor with\
shape (batch_size, num_proposals, 4), the last
dimension means
[img_width,img_height, img_width, img_height].
reduction_override (str, optional): The reduction
method used to override the original reduction
method of the loss. Options are "none",
"mean" and "sum". Defaults to None,
Returns:
dict[str, Tensor]: Dictionary of loss components
"""
losses = dict()
bg_class_ind = self.num_classes
# note in spare rcnn num_gt == num_pos
pos_inds = (labels >= 0) & (labels < bg_class_ind)
num_pos = pos_inds.sum().float()
avg_factor = reduce_mean(num_pos)
if cls_score is not None:
if cls_score.numel() > 0:
losses['loss_cls'] = self.loss_cls(
cls_score,
labels,
label_weights,
avg_factor=avg_factor,
reduction_override=reduction_override)
losses['pos_acc'] = accuracy(cls_score[pos_inds],
labels[pos_inds])
if bbox_pred is not None:
# 0~self.num_classes-1 are FG, self.num_classes is BG
# do not perform bounding box regression for BG anymore.
if pos_inds.any():
pos_bbox_pred = bbox_pred.reshape(bbox_pred.size(0),
4)[pos_inds.type(torch.bool)]
imgs_whwh = imgs_whwh.reshape(bbox_pred.size(0),
4)[pos_inds.type(torch.bool)]
losses['loss_bbox'] = self.loss_bbox(
pos_bbox_pred / imgs_whwh,
bbox_targets[pos_inds.type(torch.bool)] / imgs_whwh,
bbox_weights[pos_inds.type(torch.bool)],
avg_factor=avg_factor)
losses['loss_iou'] = self.loss_iou(
pos_bbox_pred,
bbox_targets[pos_inds.type(torch.bool)],
bbox_weights[pos_inds.type(torch.bool)],
avg_factor=avg_factor)
else:
losses['loss_bbox'] = bbox_pred.sum() * 0
losses['loss_iou'] = bbox_pred.sum() * 0
return losses
def loss(self,
cls_scores,
bbox_preds,
objectnesses,
gt_bboxes,
gt_labels,
img_metas,
gt_bboxes_ignore=None):
"""Compute loss of the head.
Args:
cls_scores (list[Tensor]): Box scores for each scale level,
each is a 4D-tensor, the channel number is
num_points * num_classes.
bbox_preds (list[Tensor]): Box energies / deltas for each scale
level, each is a 4D-tensor, the channel number is
num_points * 4.
objectnesses (list[Tensor]): objectness for each scale level, each
is a 4D-tensor, the channel number is num_points * 1.
gt_bboxes (list[Tensor]): Ground truth bboxes for each image with
shape (num_gts, 4) in [tl_x, tl_y, br_x, br_y] format.
gt_labels (list[Tensor]): class indices corresponding to each box
img_metas (list[dict]): Meta information of each image, e.g.,
image size, scaling factor, etc.
gt_bboxes_ignore (None | list[Tensor]): specify which bounding
boxes can be ignored when computing the loss.
Returns:
dict[str, Tensor]: A dictionary of loss components.
"""
assert len(cls_scores) == len(bbox_preds) == len(objectnesses)
all_num_gt = sum([len(item) for item in gt_bboxes])
featmap_sizes = [featmap.size()[-2:] for featmap in cls_scores]
all_level_points = self.prior_generator.grid_priors(
featmap_sizes,
dtype=bbox_preds[0].dtype,
device=bbox_preds[0].device)
inside_gt_bbox_mask_list, bbox_targets_list = self.get_targets(
all_level_points, gt_bboxes)
center_prior_weight_list = []
temp_inside_gt_bbox_mask_list = []
for gt_bboxe, gt_label, inside_gt_bbox_mask in zip(
gt_bboxes, gt_labels, inside_gt_bbox_mask_list):
center_prior_weight, inside_gt_bbox_mask = \
self.center_prior(all_level_points, gt_bboxe, gt_label,
inside_gt_bbox_mask)
center_prior_weight_list.append(center_prior_weight)
temp_inside_gt_bbox_mask_list.append(inside_gt_bbox_mask)
inside_gt_bbox_mask_list = temp_inside_gt_bbox_mask_list
mlvl_points = torch.cat(all_level_points, dim=0)
bbox_preds = levels_to_images(bbox_preds)
cls_scores = levels_to_images(cls_scores)
objectnesses = levels_to_images(objectnesses)
reg_loss_list = []
ious_list = []
num_points = len(mlvl_points)
for bbox_pred, encoded_targets, inside_gt_bbox_mask in zip(
bbox_preds, bbox_targets_list, inside_gt_bbox_mask_list):
temp_num_gt = encoded_targets.size(1)
expand_mlvl_points = mlvl_points[:, None, :].expand(
num_points, temp_num_gt, 2).reshape(-1, 2)
encoded_targets = encoded_targets.reshape(-1, 4)
expand_bbox_pred = bbox_pred[:, None, :].expand(
num_points, temp_num_gt, 4).reshape(-1, 4)
decoded_bbox_preds = self.bbox_coder.decode(
expand_mlvl_points, expand_bbox_pred)
decoded_target_preds = self.bbox_coder.decode(
expand_mlvl_points, encoded_targets)
with torch.no_grad():
ious = bbox_overlaps(decoded_bbox_preds,
decoded_target_preds,
is_aligned=True)
ious = ious.reshape(num_points, temp_num_gt)
if temp_num_gt:
ious = ious.max(dim=-1, keepdim=True).values.repeat(
1, temp_num_gt)
else:
ious = ious.new_zeros(num_points, temp_num_gt)
ious[~inside_gt_bbox_mask] = 0
ious_list.append(ious)
loss_bbox = self.loss_bbox(decoded_bbox_preds,
decoded_target_preds,
weight=None,
reduction_override='none')
reg_loss_list.append(loss_bbox.reshape(num_points, temp_num_gt))
cls_scores = [item.sigmoid() for item in cls_scores]
objectnesses = [item.sigmoid() for item in objectnesses]
pos_loss_list, = multi_apply(self.get_pos_loss_single, cls_scores,
objectnesses, reg_loss_list, gt_labels,
center_prior_weight_list)
pos_avg_factor = reduce_mean(
bbox_pred.new_tensor(all_num_gt)).clamp_(min=1)
pos_loss = sum(pos_loss_list) / pos_avg_factor
neg_loss_list, = multi_apply(self.get_neg_loss_single, cls_scores,
objectnesses, gt_labels, ious_list,
inside_gt_bbox_mask_list)
neg_avg_factor = sum(item.data.sum()
for item in center_prior_weight_list)
neg_avg_factor = reduce_mean(neg_avg_factor).clamp_(min=1)
neg_loss = sum(neg_loss_list) / neg_avg_factor
center_loss = []
for i in range(len(img_metas)):
if inside_gt_bbox_mask_list[i].any():
center_loss.append(
len(gt_bboxes[i]) /
center_prior_weight_list[i].sum().clamp_(min=EPS))
# when width or height of gt_bbox is smaller than stride of p3
else:
center_loss.append(center_prior_weight_list[i].sum() * 0)
center_loss = torch.stack(center_loss).mean() * self.center_loss_weight
# avoid dead lock in DDP
if all_num_gt == 0:
pos_loss = bbox_preds[0].sum() * 0
dummy_center_prior_loss = self.center_prior.mean.sum(
) * 0 + self.center_prior.sigma.sum() * 0
center_loss = objectnesses[0].sum() * 0 + dummy_center_prior_loss
loss = dict(loss_pos=pos_loss,
loss_neg=neg_loss,
loss_center=center_loss)
return loss