def _network(self):
# Convert scores to probabilities
self.rpn_cls_prob = rpn_softmax(self.rpn_cls_score)
# Determine best proposals
blobs = proposal_layer(rpn_bbox_cls_prob=self.rpn_cls_prob, rpn_bbox_pred=self.rpn_bbox_pred, im_dims=self.im_dims, cfg_key=self.key, _feat_stride=self.rpn_net._feat_stride, anchor_scales=self.flags['anchor_scales'])
# Calculate targets for proposals
self.rois, self.labels, self.bbox_targets, self.bbox_inside_weights, self.bbox_outside_weights = \
proposal_target_layer(rpn_rois=blobs, gt_boxes=self.gt_boxes,_num_classes=self.flags['num_classes'])
def _network(self):
# There shouldn't be any gt_boxes if in evaluation mode
if self.eval_mode is True:
assert self.gt_boxes is None, \
'Evaluation mode should not have ground truth boxes (or else what are you detecting for?)'
with tf.variable_scope('roi_proposal'):
# Convert scores to probabilities 转换得分到概率
self.rpn_cls_prob = rpn_softmax(self.rpn_cls_score)
# Determine best proposals
key = 'TRAIN' if self.eval_mode is False else 'TEST'
self.blobs = proposal_layer(rpn_bbox_cls_prob=self.rpn_cls_prob,
rpn_bbox_pred=self.rpn_bbox_pred,
im_dims=self.im_dims,
cfg_key=key,
_feat_stride=self.rpn_net._feat_stride,
anchor_scales=self.anchor_scales)
def _network(self):
# There shouldn't be any gt_boxes if in evaluation mode
if self.eval_mode is True:
assert self.gt_boxes is None, \
'Evaluation mode should not have ground truth boxes (or else what are you detecting for?)'
with tf.variable_scope('roi_proposal'):
# Convert scores to probabilities
self.rpn_cls_prob = rpn_softmax(self.rpn_cls_score)
# Determine best proposals
key = 'TRAIN' if self.eval_mode is False else 'TEST'
self.blobs = proposal_layer(rpn_bbox_cls_prob=self.rpn_cls_prob, rpn_bbox_pred=self.rpn_bbox_pred,
im_dims=self.im_dims, cfg_key=key, _feat_stride=self.rpn_net._feat_stride,
anchor_scales=self.anchor_scales)
if self.eval_mode is False:
# Calculate targets for proposals
self.rois, self.labels, self.bbox_targets, self.bbox_inside_weights, self.bbox_outside_weights = \
proposal_target_layer(rpn_rois=self.blobs, gt_boxes=self.gt_boxes,
_num_classes=self.num_classes)
def _network(self):
# There shouldn't be any gt_boxes if in evaluation mode
if self.eval_mode == True:
assert self.gt_boxes == None, \
'Evaluation mode should not have ground truth boxes (or else what are you detecting for?)'
# Convert scores to probabilities
self.rpn_cls_prob = rpn_softmax(self.rpn_cls_score)
# Determine best proposals
key = 'TRAIN' if self.eval_mode is False else 'TEST'
self.blobs = proposal_layer(rpn_bbox_cls_prob=self.rpn_cls_prob,
rpn_bbox_pred=self.rpn_bbox_pred,
im_dims=self.im_dims,
cfg_key=key,
_feat_stride=self.rpn_net._feat_stride,
anchor_scales=self.flags['anchor_scales'])
if self.eval_mode == False:
# Calculate targets for proposals
self.rois, self.labels, self.bbox_targets, self.bbox_inside_weights, self.bbox_outside_weights = \
proposal_target_layer(rpn_rois=self.blobs, gt_boxes=self.gt_boxes,_num_classes=self.flags['num_classes'])