Datasets:
celinelee
/
python_fn_calls

Dataset card Data Studio Files
xet
 Community
1
code
stringlengths
3
6.57k
tf.variable_scope(scope, values=[preprocessed_inputs])
self._extract_proposal_features(preprocessed_inputs, scope)
_extract_proposal_features(self, preprocessed_inputs, scope)
extract_box_classifier_features(self, proposal_feature_maps, scope)
self._extract_box_classifier_features(proposal_feature_maps, scope)
_extract_box_classifier_features(self, proposal_feature_maps, scope)
names (to load from a checkpoint)
tf.global_variables()
variable.op.name.startswith(scope_name)
variable.op.name.replace(scope_name + '/', '')
FasterRCNNMetaArchOverrideRPN(model.DetectionModel)
Network (RPN)
CNN (RPN)
field. (This should typically be set to 1)
loss (RPN loss)
Network (RPN)
Union (IOU)
Suppression (NMS)
Network (RPN)
Suppression (NMS)
Network (RPN)
size (width and height are set to be the same)
optional (kwarg)
inputs (with all other inputs already set)
and (optionally)
nonlinearity (that takes tensors as inputs and returns tensors)
object (can be None)
boolean (default: True)
TODO(rathodv)
print("Running FasterRCNN with overriden RPN")
super(FasterRCNNMetaArchOverrideRPN, self)
__init__(num_classes=num_classes)
if (number_of_stages==1)
ValueError('Number of stages = 1 is not allowed for overriden RPN proposals')
os.path.join(first_stage_proposals_path, '')
data_util.read_xml_batch(first_stage_proposals_path)
data_util.xml_to_numpy(None, xml_root)
print("Shape of overriding proposals",self.proposals.shape)
losses.WeightedSmoothL1LocalizationLoss()
losses.WeightedSoftmaxClassificationLoss()
losses.WeightedSmoothL1LocalizationLoss()
losses.WeightedSigmoidClassificationLoss()
ValueError('Number of stages should be a value in {1, 2, 3}.')
first_stage_feature_extractor_scope(self)
second_stage_feature_extractor_scope(self)
first_stage_box_predictor_scope(self)
second_stage_box_predictor_scope(self)
max_num_proposals(self)
proposals (to pad to)
anchors(self)
RuntimeError('anchors have not been constructed yet!')
isinstance(self._anchors, box_list.BoxList)
RuntimeError('anchors should be a BoxList object, but is not.')
preprocess(self, inputs)
preprocessing (e.g., scaling pixel values to be in [-1, 1])
ValueError('`preprocess` expects a tf.float32 tensor')
tf.name_scope('Preprocessor')
return (self._feature_extractor.preprocess(resized_inputs)
_compute_clip_window(self, image_shapes)
at (0, 0)
tf.to_float(tf.stack([tf.zeros_like(clip_heights)
tf.zeros_like(clip_heights)
predict(self, preprocessed_inputs, true_image_shapes)
predictions (un-postprocessed)
predictions (logits)
predictions (at class index 0)
RPN (in absolute coordinates)
mode. (and if number_of_stages > 1)
predicted (final)
predictions (logits)
predictions (at class index 0)
self._extract_rpn_feature_maps(preprocessed_inputs)
print("Predict running")
tf.shape(preprocessed_inputs)
self._predict_rpn_proposals(rpn_box_predictor_features)
tf.to_float(tf.stack([0, 0, image_shape[1], image_shape[2]])
self._anchors.get()
self._anchors.get()
_image_batch_shape_2d(self, image_batch_shape_1d)
tf.tile(tf.expand_dims(image_batch_shape_1d[1:], 0)
predictions (logits)
predictions (at class index 0)
predicted (final)
predictions (logits)
predictions (at class index 0)
self._image_batch_shape_2d(image_shape)
tf.Print(proposal_boxes_normalized, [], message=("original size= " + str(proposal_boxes_normalized.shape[1])
tf.constant(self.proposals, dtype='float32')
predicted (final)
predictions (logits)
predictions (at class index 0)
self._image_batch_shape_2d(image_shape)
tf.Print(proposal_boxes_normalized, [], message=("original size= " + str(proposal_boxes_normalized.shape[1])
normalize_boxes(args)
box_list.BoxList(proposal_boxes_per_image)
get()
to_absolute_boxes(args)
box_list.BoxList(proposal_boxes_per_image)
get()
tf.constant(self.proposals, dtype='float32')