def forward(self, bottom, top):
# Algorithm:
#
# for each (H, W) location i
# generate A anchor boxes centered on cell i
# apply predicted bbox deltas at cell i to each of the A anchors
# clip predicted boxes to image
# remove predicted boxes with either height or width < threshold
# sort all (proposal, score) pairs by score from highest to lowest
# take top pre_nms_topN proposals before NMS
# apply NMS with threshold 0.7 to remaining proposals
# take after_nms_topN proposals after NMS
# return the top proposals (-> RoIs top, scores top)
assert bottom[0].data.shape[0] == 1, \
'Only single item batches are supported'
cfg_key = 'TEST' #str(self.phase) # either 'TRAIN' or 'TEST'
pre_nms_topN = cfg[cfg_key].RPN_PRE_NMS_TOP_N
post_nms_topN = cfg[cfg_key].RPN_POST_NMS_TOP_N
nms_thresh = cfg[cfg_key].RPN_NMS_THRESH
min_size = cfg[cfg_key].RPN_MIN_SIZE
# the first set of _num_anchors channels are bg probs
# the second set are the fg probs, which we want
scores = bottom[0].data[:, self._num_anchors:, :, :] # ??????
bbox_deltas = bottom[1].data
im_info = bottom[2].data[0, :]
if DEBUG:
print 'im_size: ({}, {})'.format(im_info[0], im_info[1])
print 'scale: {}'.format(im_info[2])
# 1. Generate proposals from bbox deltas and shifted anchors
height, width = scores.shape[-2:]
if DEBUG:
print 'score map size: {}'.format(scores.shape)
# Enumerate all shifts
shift_x = np.arange(0, width) * self._feat_stride
shift_y = np.arange(0, height) * self._feat_stride
shift_x, shift_y = np.meshgrid(shift_x, shift_y)
# ctr_x, ctr_y shift, the rest 3 colomns filled zero
shifts = np.vstack(
(shift_x.ravel(), shift_y.ravel(), np.zeros(
(3, width * height)))).transpose()
# Enumerate all shifted anchors:
#
# add A anchors (1, A, 5) to
# cell K shifts (K, 1, 5) to get
# shift anchors (K, A, 5)
# reshape to (K*A, 5) shifted anchors
A = self._num_anchors
K = shifts.shape[0]
anchors = self._anchors.reshape(
(1, A, self._bbox_para_num)) + shifts.reshape(
(1, K, self._bbox_para_num)).transpose((1, 0, 2)) # D
anchors = anchors.reshape((K * A, self._bbox_para_num)) # D
#print anchors
# Transpose and reshape predicted bbox transformations to get them
# into the same order as the anchors:
#
# bbox deltas will be (1, 5 * A, H, W) format
# transpose to (1, H, W, 5 * A)
# reshape to (1 * H * W * A, 5) where rows are ordered by (h, w, a)
# in slowest to fastest order
# [ctr_x, ctr_y, height, width, angle]
bbox_deltas = bbox_deltas.transpose((0, 2, 3, 1)).reshape((-1, 5)) # D
# Same story for the scores:
#
# scores are (1, A, H, W) format
# transpose to (1, H, W, A)
# reshape to (1 * H * W * A, 1) where rows are ordered by (h, w, a)
scores = scores.transpose((0, 2, 3, 1)).reshape((-1, 1))
# Convert anchors into proposals via bbox transformations
#print bbox_deltas[:, 4]
#bbox_deltas[:, 4] = 0
proposals = rbbox_transform_inv(anchors, bbox_deltas) # D
# 2. clip predicted boxes to image
# proposals = clip_boxes(proposals, im_info[:2]) # TODO
# 3. remove predicted boxes with either height or width < threshold
# (NOTE: convert min_size to input image scale stored in im_info[2])
keep = _filter_boxes(proposals, min_size * im_info[2]) # D
proposals = proposals[keep, :]
scores = scores[keep]
# 4. sort all (proposal, score) pairs by score from highest to lowest
# 5. take top pre_nms_topN (e.g. 6000)
order = scores.ravel().argsort()[::-1]
if pre_nms_topN > 0:
order = order[:pre_nms_topN]
proposals = proposals[order, :]
scores = scores[order]
###
#anchors = anchors[keep,:]
#anchors = anchors[order,:]
###
# 6. apply nms (e.g. threshold = 0.7)
# 7. take after_nms_topN (e.g. 300)
# 8. return the top proposals (-> RoIs top)
import time
tic = time.time()
keep = rotate_cpu_nms(np.hstack((proposals, scores)), nms_thresh,
cfg.GPU_ID) # D
#print time.time() - tic
if post_nms_topN > 0:
keep = keep[:post_nms_topN]
proposals = proposals[keep, :]
scores = scores[keep]
#anchors = anchors[keep,:]
#for i in range(300):
# print anchors[i]
# print proposals[i]
# print scores[i]
# Output rois blob
# Our RPN implementation only supports a single input image, so all
# batch inds are 0
batch_inds = np.zeros((proposals.shape[0], 1), dtype=np.float32)
blob = np.hstack((batch_inds, proposals.astype(np.float32,
copy=False)))
top[0].reshape(*(blob.shape))
top[0].data[...] = blob
# [Optional] output scores blob
if len(top) > 1:
top[1].reshape(*(scores.shape))
top[1].data[...] = scores
def forward(self, bottom, top):
# Algorithm:
#
# for each (H, W) location i
# generate A anchor boxes centered on cell i
# apply predicted bbox deltas at cell i to each of the A anchors
# clip predicted boxes to image
# remove predicted boxes with either height or width < threshold
# sort all (proposal, score) pairs by score from highest to lowest
# take top pre_nms_topN proposals before NMS
# apply NMS with threshold 0.7 to remaining proposals
# take after_nms_topN proposals after NMS
# return the top proposals (-> RoIs top, scores top)
assert bottom[0].data.shape[0] == 1, \
'Only single item batches are supported'
cfg_key = str(self.phase) # either 'TRAIN' or 'TEST'
pre_nms_topN = cfg[cfg_key].RPN_PRE_NMS_TOP_N
post_nms_topN = cfg[cfg_key].RPN_POST_NMS_TOP_N
nms_thresh = cfg[cfg_key].RPN_NMS_THRESH
min_size = cfg[cfg_key].RPN_MIN_SIZE
# the first set of _num_anchors channels are bg probs
# the second set are the fg probs, which we want
scores = bottom[0].data[:, self._num_anchors:, :, :] # ??????
bbox_deltas = bottom[1].data
im_info = bottom[2].data[0, :]
if DEBUG:
print 'im_size: ({}, {})'.format(im_info[0], im_info[1])
print 'scale: {}'.format(im_info[2])
# 1. Generate proposals from bbox deltas and shifted anchors
height, width = scores.shape[-2:]
if DEBUG:
print 'score map size: {}'.format(scores.shape)
# Enumerate all shifts
shift_x = np.arange(0, width) * self._feat_stride
shift_y = np.arange(0, height) * self._feat_stride
shift_x, shift_y = np.meshgrid(shift_x, shift_y)
# ctr_x, ctr_y shift, the rest 3 colomns filled zero
shifts = np.vstack((shift_x.ravel(), shift_y.ravel(),
np.zeros((3, width * height)))).transpose()
# Enumerate all shifted anchors:
#
# add A anchors (1, A, 5) to
# cell K shifts (K, 1, 5) to get
# shift anchors (K, A, 5)
# reshape to (K*A, 5) shifted anchors
A = self._num_anchors
K = shifts.shape[0]
anchors = self._anchors.reshape((1, A, self._bbox_para_num)) + shifts.reshape((1, K, self._bbox_para_num)).transpose((1, 0, 2)) # D
anchors = anchors.reshape((K * A, self._bbox_para_num)) # D
#print anchors
# Transpose and reshape predicted bbox transformations to get them
# into the same order as the anchors:
#
# bbox deltas will be (1, 5 * A, H, W) format
# transpose to (1, H, W, 5 * A)
# reshape to (1 * H * W * A, 5) where rows are ordered by (h, w, a)
# in slowest to fastest order
# [ctr_x, ctr_y, height, width, angle]
bbox_deltas = bbox_deltas.transpose((0, 2, 3, 1)).reshape((-1, 5)) # D
# Same story for the scores:
#
# scores are (1, A, H, W) format
# transpose to (1, H, W, A)
# reshape to (1 * H * W * A, 1) where rows are ordered by (h, w, a)
scores = scores.transpose((0, 2, 3, 1)).reshape((-1, 1))
# Convert anchors into proposals via bbox transformations
#print bbox_deltas[:, 4]
#bbox_deltas[:, 4] = 0
proposals = rbbox_transform_inv(anchors, bbox_deltas) # D
# 2. clip predicted boxes to image
# proposals = clip_boxes(proposals, im_info[:2]) # TODO
# 3. remove predicted boxes with either height or width < threshold
# (NOTE: convert min_size to input image scale stored in im_info[2])
keep = _filter_boxes(proposals, min_size * im_info[2]) # D
proposals = proposals[keep, :]
scores = scores[keep]
# 4. sort all (proposal, score) pairs by score from highest to lowest
# 5. take top pre_nms_topN (e.g. 6000)
order = scores.ravel().argsort()[::-1]
if pre_nms_topN > 0:
order = order[:pre_nms_topN]
proposals = proposals[order, :]
scores = scores[order]
###
#anchors = anchors[keep,:]
#anchors = anchors[order,:]
###
# 6. apply nms (e.g. threshold = 0.7)
# 7. take after_nms_topN (e.g. 300)
# 8. return the top proposals (-> RoIs top)
import time
tic = time.time()
keep = rotate_cpu_nms(np.hstack((proposals, scores)), nms_thresh,cfg.GPU_ID) # D
#print time.time() - tic
if post_nms_topN > 0:
keep = keep[:post_nms_topN]
proposals = proposals[keep, :]
scores = scores[keep]
#anchors = anchors[keep,:]
#for i in range(300):
# print anchors[i]
# print proposals[i]
# print scores[i]
# Output rois blob
# Our RPN implementation only supports a single input image, so all
# batch inds are 0
batch_inds = np.zeros((proposals.shape[0], 1), dtype=np.float32)
blob = np.hstack((batch_inds, proposals.astype(np.float32, copy=False)))
top[0].reshape(*(blob.shape))
top[0].data[...] = blob
# [Optional] output scores blob
if len(top) > 1:
top[1].reshape(*(scores.shape))
top[1].data[...] = scores
(3, width * height)))).transpose()
A = _num_anchors
K = shifts.shape[0]
anchors = _anchors.reshape((1, A, _bbox_para_num)) + shifts.reshape(
(1, K, _bbox_para_num)).transpose((1, 0, 2)) # D
anchors = anchors.reshape((K * A, _bbox_para_num)) # D
print "A", A, "K", K
bbox_deltas = bbox_deltas.transpose((0, 2, 3, 1)).reshape((-1, 5)) # D
scores = scores.transpose((0, 2, 3, 1)).reshape((-1, 1))
proposals = rbbox_transform_inv(anchors, bbox_deltas)
# proposals = clip_boxes(proposals, im_info[:2]) # TODO
keep = _filter_boxes(proposals, min_size * im_info[2]) # D
proposals = proposals[keep, :]
scores = scores[keep]
order = scores.ravel().argsort()[::-1]
if pre_nms_topN > 0:
order = order[:pre_nms_topN]
proposals = proposals[order, :]
scores = scores[order]
print scores.shape
print proposals.shape
anchors = _anchors.reshape((1, A, _bbox_para_num)) + shifts.reshape((1, K, _bbox_para_num)).transpose((1, 0, 2)) # D
anchors = anchors.reshape((K * A, _bbox_para_num)) # D
print "A", A, "K", K
bbox_deltas = bbox_deltas.transpose((0, 2, 3, 1)).reshape((-1, 5)) # D
scores = scores.transpose((0, 2, 3, 1)).reshape((-1, 1))
proposals = rbbox_transform_inv(anchors, bbox_deltas)
# proposals = clip_boxes(proposals, im_info[:2]) # TODO
keep = _filter_boxes(proposals, min_size * im_info[2]) # D
proposals = proposals[keep, :]
scores = scores[keep]
order = scores.ravel().argsort()[::-1]
if pre_nms_topN > 0:
order = order[:pre_nms_topN]
proposals = proposals[order, :]
scores = scores[order]