def bboxes_encode(self, labels, bboxes, anchors,
scope='ssd_bboxes_encode'):
"""Encode labels and bounding boxes.
"""
return ssd_common.tf_ssd_bboxes_encode(
labels, bboxes, anchors,
matching_threshold=0.5,
prior_scaling=self.params.prior_scaling,
scope=scope)
def bboxes_encode(self, labels, bboxes, anchors, scope=None):
"""Encode labels and bounding boxes.
"""
return ssd_common.tf_ssd_bboxes_encode(
labels, bboxes, anchors,
self.params.num_classes,
self.params.no_annotation_label,
ignore_threshold=0.5,
prior_scaling=self.params.prior_scaling,
scope=scope)
def bboxes_encode(self, labels, bboxes, anchors,
scope=None):
"""Encode labels and bounding boxes.
"""
return ssd_common.tf_ssd_bboxes_encode(
labels, bboxes, anchors,
self.params.num_classes,
self.params.no_annotation_label,
ignore_threshold=0.5,
prior_scaling=self.params.prior_scaling,
scope=scope)
def bboxes_encode(self, labels, bboxes, anchors, #编码，用于将标签信息，真实目标信息和锚点框信息编码在一起；得到预测真实框到参考框的转换值
scope=None):
"""Encode labels and bounding boxes.
"""
return ssd_common.tf_ssd_bboxes_encode(
labels, bboxes, anchors,
self.params.num_classes,
self.params.no_annotation_label, #未标注的标签（应该代表背景）
ignore_threshold=0.5, #IOU筛选阈值
prior_scaling=self.params.prior_scaling, #特征图目标与参考框间的尺寸缩放（0.1,0.1,0.2,0.2）
scope=scope)
def bboxes_encode(self,
labels,
bboxes,
anchors,
scope='ssd_bboxes_encode'):
"""Encode labels and bounding boxes.
"""
return ssd_common.tf_ssd_bboxes_encode(
labels,
bboxes,
anchors,
matching_threshold=0.5,
prior_scaling=self.params.prior_scaling,
scope=scope)
def bboxes_encode(self,
labels,
bboxes,
anchors,
positive_threshold=0.5,
ignore_threshold=0.3,
scope=None):
"""Encode labels and bounding boxes.
"""
return ssd_common.tf_ssd_bboxes_encode(
labels,
bboxes,
anchors,
self.params.num_classes,
self.params.img_shape,
self.params.allowed_borders,
self.params.no_annotation_label,
positive_threshold=positive_threshold,
ignore_threshold=ignore_threshold,
prior_scaling=self.params.prior_scaling,
scope=scope)
reuse = True if 'ssd' in locals() else None
params = ssd_vgg_300.SSDNet.default_params
ssd = ssd_vgg_300.SSDNet(params)
with slim.arg_scope(ssd.arg_scope(weight_decay=0.0005)):
predictions, localisations, logits, end_points = ssd.net(image_4d,
is_training=False,
reuse=reuse)
# SSD default anchor boxes.
img_shape = out_shape
layers_anchors = ssd.anchors(img_shape, dtype=np.float32)
# In[41]:
# Targets encoding.
target_labels, target_localizations, target_scores = ssd_common.tf_ssd_bboxes_encode(
labels, bboxes_pre, layers_anchors, 8, 8)
# In[43]:
# Initialize variables.
init_op = tf.global_variables_initializer()
sess.run(init_op)
# Restore SSD model.
saver = tf.train.Saver()
saver.restore(sess, ckpt_filename)
# In[45]:
# Run model.
[
rimg, rpredictions, rlocalisations, glabels, gbboxes, rbbox_img, rt_labels,
# SSD construction.
with slim.arg_scope(ssd.arg_scope(weight_decay=0.0005)):
predictions, localisations, logits, end_points = ssd.net(image_4d,
is_training=False,
reuse=reuse)
# SSD default anchor boxes.
img_shape = out_shape
layers_anchors = ssd.anchors(img_shape, dtype=np.float32)
for k in sorted(end_points.keys()):
print(k, end_points[k].get_shape())
# Targets encoding.
target_labels, target_localizations, target_scores = \
ssd_common.tf_ssd_bboxes_encode(labels, bboxes_pre, layers_anchors,
num_classes=params.num_classes, no_annotation_label=params.no_annotation_label)
nms_threshold = 0.5
# Output decoding.
localisations = ssd.bboxes_decode(localisations, layers_anchors)
tscores, tbboxes = ssd.detected_bboxes(predictions,
localisations,
select_threshold=0.8,
nms_threshold=0.5)
# Initialize variables.
init_op = tf.global_variables_initializer()
isess.run(init_op)
# Restore SSD model.
saver = tf.train.Saver()
def bboxes_encode(self, labels, bboxes, anchors, scope=None):
"""
Encode labels and bounding boxes.
"""
return ssd_common.tf_ssd_bboxes_encode()
