def _compute_targets(ex_rois, gt_rois, gt_rrois, gt_quadrois, labels):
"""Compute bounding-box regression targets for an image."""
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 4
assert gt_rrois.shape[1] == 5
assert gt_quadrois.shape[1] == 8
targets = bbox_transform(ex_rois, gt_rois)
rtargets = rbox_transform(ex_rois, gt_rrois)
quadtargets = quadbox_transform(ex_rois, gt_quadrois)
if cfg.TRAIN.BBOX_NORMALIZE_TARGETS_PRECOMPUTED:
# Optionally normalize targets by a precomputed mean and stdev
targets = ((targets - np.array(cfg.TRAIN.BBOX_NORMALIZE_MEANS))
/ np.array(cfg.TRAIN.BBOX_NORMALIZE_STDS))
if cfg.TRAIN.RBOX_NORMALIZE_TARGETS_PRECOMPUTED:
# Optionally normalize targets by a precomputed mean and stdev
rtargets = ((rtargets - np.array(cfg.TRAIN.RBOX_NORMALIZE_MEANS))
/ np.array(cfg.TRAIN.RBOX_NORMALIZE_STDS))
if cfg.TRAIN.QUADBOX_NORMALIZE_TARGETS_PRECOMPUTED:
# Optionally normalize targets by a precomputed mean and stdev
quadtargets = ((quadtargets - np.array(cfg.TRAIN.QUADBOX_NORMALIZE_MEANS))
/ np.array(cfg.TRAIN.QUADBOX_NORMALIZE_STDS))
return np.hstack(
(labels[:, np.newaxis], targets)).astype(np.float32, copy=False), \
np.hstack(
(labels[:, np.newaxis], rtargets)).astype(np.float32, copy=False), \
np.hstack(
(labels[:, np.newaxis], quadtargets)).astype(np.float32, copy=False)
def _compute_targets(ex_rois, gt_rois):
# 计算一张照片的边界框回归目标
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 5
return bbox_transform(ex_rois, gt_rois[:, :4]).astype(np.float32,
copy=False)
def _compute_targets(ex_rois, gt_rois): """Compute bounding-box regression targets for an image.""" assert ex_rois.shape[0] == gt_rois.shape[0] assert ex_rois.shape[1] == 4 assert gt_rois.shape[1] == 5 return bbox_transform(torch.from_numpy(ex_rois), torch.from_numpy(gt_rois[:, :4])).numpy()
def _compute_targets(ex_rois, gt_rois): """Compute bounding-box regression targets for an image.""" assert ex_rois.shape[0] == gt_rois.shape[0] assert ex_rois.shape[1] == 4 assert gt_rois.shape[1] == 5 return bbox_transform(ex_rois, gt_rois[:, :4]).astype(np.float32, copy=False)
def _compute_targets(ex_rois, gt_rois): """Compute bounding-box regression targets for an image.""" assert ex_rois.shape[0] == gt_rois.shape[0] assert ex_rois.shape[1] == 4 assert gt_rois.shape[1] == 5 return bbox_transform(ex_rois, gt_rois[:, :4]).astype(np.float32, copy=False)
def _compute_targets(ex_rois, gt_rois):
"""Compute bounding-box regression targets for an image."""
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 5
return bbox_transform(
torch.from_numpy(ex_rois), torch.from_numpy(gt_rois[:, :4])).numpy()
def _compute_targets(ex_rois,gt_rois,labels):
# 计算图像的边界框回归目标
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 4
targets = bbox_transform(ex_rois,gt_rois)
if cfg.TRAIN.BBOX_NORMALIZE_TARGETS_PRECOMPUTED:
targets = ((targets-np.array(cfg.TRAIN.BBOX_NORMALIZE_MEANS))/np.array(cfg.TRAIN.BBOX_NORMALIZE_STDS))
return np.hstack((labels[:,np.newaxis],targets)).astype(np.float32,copy=False)
def _compute_targets(ex_rois, gt_rois):
"""Compute bounding-box regression targets for an image."""
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 5
targets = bbox_transform(ex_rois, gt_rois)
return targets
def _compute_targets(ex_rois, gt_rois): """Compute bounding-box regression targets for an image.""" ## 要求anchor与对应匹配最好GT个数相同 assert ex_rois.shape[0] == gt_rois.shape[0] ## 要有anchor左上角与右下角坐标，有4个元素 assert ex_rois.shape[1] == 4 ## GT有标签位，所以为5个 assert gt_rois.shape[1] == 5 ## 返回一个用于anchor回归成target的包含每个anchor回归值(dx、dy、dw、dh)的array return bbox_transform(ex_rois, gt_rois[:, :4]).astype(np.float32, copy=False)
def _compute_targets(ex_rois, gt_rois):
"""Compute bounding-box regression targets for an image."""
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 5
targets = bbox_transform(ex_rois, gt_rois)
return targets
def _compute_targets(rois, overlaps, labels):
"""Compute bounding-box regression targets for an image."""
# We are sampling relations from fg rois, hence each
# fg box must be assigned to an gt box
assert(cfg.TRAIN.FG_THRESH >= cfg.TRAIN.BBOX_THRESH)
# Indices of ground-truth ROIs
gt_inds = np.where(overlaps == 1)[0]
if len(gt_inds) == 0:
# Bail if the image has no ground-truth ROIs
return np.zeros((rois.shape[0], 5), dtype=np.float32)
else:
# sanity check
assert(gt_inds[0] == 0)
for i in range(1, len(gt_inds)):
assert(gt_inds[i] - gt_inds[i-1] == 1)
# Indices of examples for which we try to make predictions
ex_inds = np.where(overlaps >= cfg.TRAIN.BBOX_THRESH)[0]
# Get IoU overlap between each ex ROI and gt ROI
ex_gt_overlaps = bbox_overlaps(
np.ascontiguousarray(rois[ex_inds, :], dtype=np.float),
np.ascontiguousarray(rois[gt_inds, :], dtype=np.float))
# Find which gt ROI each ex ROI has max overlap with:
# this will be the ex ROI's gt target
gt_assignment = ex_gt_overlaps.argmax(axis=1)
# guarding against the case where a gt box doesn't get assigned to itself
gt_to_ex_inds = [np.where(ex_inds == g)[0][0] for g in gt_inds]
for i, g in enumerate(gt_to_ex_inds):
gt_assignment[g] = gt_inds[i]
# assign rois
gt_rois = rois[gt_inds[gt_assignment], :]
ex_rois = rois[ex_inds, :]
# record target assignments for all foreground rois
fg_gt_ind_assignment = {}
for i, e in enumerate(ex_inds):
if overlaps[e] >= cfg.TRAIN.FG_THRESH:
fg_gt_ind_assignment[e] = gt_inds[gt_assignment[i]]
# check if all gt has been assigned
for g in gt_inds:
assert(g in list(fg_gt_ind_assignment.values()))
targets = np.zeros((rois.shape[0], 5), dtype=np.float32)
targets[ex_inds, 0] = labels[ex_inds]
# transfer to center and log
targets[ex_inds, 1:] = bbox_transform(ex_rois, gt_rois)
return targets, fg_gt_ind_assignment
def _compute_targets(ex_rois, gt_rois):
"""Compute bounding-box regression targets for an image."""
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 5
if isinstance(ex_rois, np.ndarray):
ex_rois = torch.from_numpy(ex_rois)
if isinstance(gt_rois, np.ndarray):
gt_rois = torch.from_numpy(gt_rois)
return bbox_transform(ex_rois, gt_rois[:, :4])
def _compute_targets(ex_rois, gt_rois):
"""Compute bounding-box regression targets for an image."""
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 4
targets = bbox_transform(ex_rois, gt_rois)
if cfg.TRAIN.BBOX_NORMALIZE_TARGETS_PRECOMPUTED:
# Optionally normalize targets by a precomputed mean and stdev
targets = ((targets - np.array(cfg.TRAIN.BBOX_NORMALIZE_MEANS)) /
np.array(cfg.TRAIN.BBOX_NORMALIZE_STDS))
return targets
def _compute_targets(ex_rois, gt_rois, labels):
"""Compute bounding-box regression targets for an image."""
# Inputs are tensor
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 4
targets = bbox_transform(ex_rois, gt_rois)
if cfg.TRAIN.BBOX_NORMALIZE_TARGETS_PRECOMPUTED:
# Optionally normalize targets by a precomputed mean and stdev
targets = ((targets - targets.new(cfg.TRAIN.BBOX_NORMALIZE_MEANS)) /
targets.new(cfg.TRAIN.BBOX_NORMALIZE_STDS))
return torch.cat([labels.unsqueeze(1), targets], 1)
def _compute_targets(ex_rois, gt_rois, labels):
"""Compute bounding-box regression targets for an image."""
# Inputs are tensor
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 4
targets = bbox_transform(ex_rois, gt_rois)
if cfg.TRAIN.BBOX_NORMALIZE_TARGETS_PRECOMPUTED:
# Optionally normalize targets by a precomputed mean and stdev
targets = ((targets - targets.new(cfg.TRAIN.BBOX_NORMALIZE_MEANS)) /
targets.new(cfg.TRAIN.BBOX_NORMALIZE_STDS))
return torch.cat([labels.unsqueeze(1), targets], 1)
def _compute_targets(ex_rois, gt_rois, labels):
"""Compute bounding-box regression targets for an image."""
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 4
targets = bbox_transform(ex_rois, gt_rois)
if cfg.TRAIN.BBOX_NORMALIZE_TARGETS_PRECOMPUTED:
# Optionally normalize targets by a precomputed mean and stdev
targets = ((targets - np.array(cfg.TRAIN.BBOX_NORMALIZE_MEANS))
/ np.array(cfg.TRAIN.BBOX_NORMALIZE_STDS))
return np.hstack(
(labels[:, np.newaxis], targets)).astype(np.float32, copy=False)
def _compute_targets(ex_rois, gt_rois, labels):
"""Compute bounding-box regression targets for an image."""
assert ex_rois.shape[0] == gt_rois.shape[
0] #confirm that the ex_rois and the gt_rois are the same shape[0],so as to compute the results from proposals to gt boxes.
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 4
targets = bbox_transform(ex_rois, gt_rois)
if cfg.TRAIN.BBOX_NORMALIZE_TARGETS_PRECOMPUTED:
# Optionally normalize targets by a precomputed mean and stdev
targets = ((targets - np.array(cfg.TRAIN.BBOX_NORMALIZE_MEANS)) /
np.array(cfg.TRAIN.BBOX_NORMALIZE_STDS))
return np.hstack((labels[:, np.newaxis], targets)).astype(np.float32,
copy=False)
def _compute_targets(ex_rois, gt_rois, labels):
"""Compute bounding-box regression targets for an image."""
# Inputs are tensor
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 4
# 返回roi相对与其匹配的gt (dx,dy,dw,dh)四个回归值，shape（len（rois），4）
targets = bbox_transform(ex_rois, gt_rois)
if cfg.TRAIN.BBOX_NORMALIZE_TARGETS_PRECOMPUTED:
# 是否归一化
# Optionally normalize targets by a precomputed mean and stdev
targets = ((targets - targets.new(cfg.TRAIN.BBOX_NORMALIZE_MEANS))
/ targets.new(cfg.TRAIN.BBOX_NORMALIZE_STDS))
# labels.unsqueeze(1) -> (128, 1)
return torch.cat(
[labels.unsqueeze(1), targets], 1) # (128, 5) 类别和4个回归值
def _compute_targets(ex_rois, gt_rois, labels):
"""Compute bounding-box regression targets for an image."""
assert ex_rois.shape[0] == gt_rois.shape[0]
assert ex_rois.shape[1] == 4
assert gt_rois.shape[1] == 4
#targets:是anchor距离gt_box的 :[dx, dy，dw, dh]
targets = bbox_transform(ex_rois, gt_rois)
#cfg.TRAIN.BBOX_NORMALIZE_TARGETS_PRECOMPUTED=True
if cfg.TRAIN.BBOX_NORMALIZE_TARGETS_PRECOMPUTED:
# Optionally normalize targets by a precomputed mean and stdev
'''
下面这条语句的作用不懂???
'''
targets = ((targets - np.array(cfg.TRAIN.BBOX_NORMALIZE_MEANS)) /
np.array(cfg.TRAIN.BBOX_NORMALIZE_STDS))
return np.hstack((labels[:, np.newaxis], targets)).astype(np.float32,
copy=False)
def forward(self, gt_boxes, gt_ellipses):
batch_size = gt_boxes.size(0)
labels = gt_boxes.new(batch_size, self._inds_inside.size(0)).fill_(-1)
overlaps = bbox_overlaps_batch(self._anchors, gt_boxes)
max_overlaps, argmax_overlaps = torch.max(overlaps, 2)
gt_max_overlaps, _ = torch.max(overlaps, 1)
# assign bg labels first so that positive labels can clobber them
labels[max_overlaps < self._cfg['TRAIN.RPN_NEGATIVE_OVERLAP']] = 0
# mark max overlap of 0, which are padded gt_boxes
gt_max_overlaps[gt_max_overlaps == 0] = 1e-5
# mark the max overlap of each ground truth
keep = torch.sum(overlaps.eq(
gt_max_overlaps.view(batch_size, 1, -1).expand_as(overlaps)), 2)
if torch.sum(keep) > 0:
labels[keep > 0] = 1
labels[max_overlaps >= self._cfg['TRAIN.RPN_POSITIVE_OVERLAP']] = 1
# subsample positive labels if we have too many
num_fg = int(self._cfg['TRAIN.RPN_FG_FRACTION'] *
self._cfg['TRAIN.RPN_BATCHSIZE'])
sum_fg = torch.sum((labels == 1).int(), 1)
sum_bg = torch.sum((labels == 0).int(), 1)
bbox_targets = gt_boxes.new(
batch_size, self._inds_inside.size(0), 4).fill_(0)
ellipse_targets = gt_ellipses.new(
batch_size, self._inds_inside.size(0), 5).fill_(0)
for i in range(batch_size):
# subsample positive labels if we have too many
if sum_fg[i] > num_fg:
fg_inds = torch.nonzero(labels[i] == 1).view(-1)
rand_num = torch.from_numpy(
np.random.permutation(fg_inds.size(0))
).type_as(gt_boxes).long()
disable_inds = fg_inds[rand_num[:fg_inds.size(0)-num_fg]]
labels[i][disable_inds] = -1
num_bg = self._cfg['TRAIN.RPN_BATCHSIZE'] - \
torch.sum((labels == 1).int(), 1)[i]
# subsample negative labels if we have too many
if sum_bg[i] > num_bg:
bg_inds = torch.nonzero(labels[i] == 0).view(-1)
rand_num = torch.from_numpy(
np.random.permutation(bg_inds.size(0))
).type_as(gt_boxes).long()
disable_inds = bg_inds[rand_num[:bg_inds.size(0)-num_bg]]
labels[i][disable_inds] = -1
bbox_targets[i] = bbox_transform(
self._anchors, gt_boxes[i][argmax_overlaps[i], :4])
ellipse_targets[i] = ellipse_transform(
self._anchors, gt_ellipses[i][argmax_overlaps[i], :8])
# map up to original set of anchors
labels = _unmap(
labels, self._total_anchors, self._inds_inside, batch_size,
fill=-1)
bbox_targets = _unmap(
bbox_targets, self._total_anchors, self._inds_inside, batch_size,
fill=0)
ellipse_targets = _unmap(
ellipse_targets, self._total_anchors, self._inds_inside,
batch_size, fill=0)
labels = labels.view(
batch_size, self._feat_height, self._feat_width, self._A, 1
).contiguous()
bbox_targets = bbox_targets.view(
batch_size, self._feat_height, self._feat_width, self._A, 4
).contiguous()
ellipse_targets = ellipse_targets.view(
batch_size, self._feat_height, self._feat_width, self._A, 5
).contiguous()
return (labels, bbox_targets, ellipse_targets)
sub_image = image[10*j:10*j+200,10*i:10*i+200,:]#[y:y+h,x:x+w]
sub_image = cv2.resize(sub_image,(32,32))
sub_image = sub_image[np.newaxis,:,:,:]
area = np.array([10*i,10*j,10*i+200,10*j+200])#[x,y]
area = area[np.newaxis,:]
# print(area)
# print(sub_image.shape)
image_batch.append(sub_image)
image_area.append(area)
image_batch = np.concatenate(image_batch,axis=0)
anchors = np.concatenate(image_area,axis=0)
# print(anchors.shape)
print(image_batch.shape)
print(image_batch.dtype)
targets = bbox_transform(anchors, label)
# print('11111111111')
# print(targets,targets.shape,targets.dtype)
print(targets.shape)
# print(targets)
overlaps = bbox_overlaps(
np.ascontiguousarray(anchors, dtype=np.float),
np.ascontiguousarray(label_sub, dtype=np.float))
# print(overlaps)
n = np.where(overlaps[:,0] > 0.7)
print(n)
face_label_1 = np.ones([231,1])
face_label_2 = np.zeros([231,1])
face_label_2[n,0]=1
face_label_1[n,0]=0