def infer_outputs(self):
# sampled rois (0, x1, y1, x2, y2)
# for CNTK the proposal shape is [4 x roisPerImage], and mirrored in Python
rois_shape = (FreeDimension, 4)
labels_shape = (FreeDimension, self._num_classes)
bbox_targets_shape = (FreeDimension, self._num_classes * 4)
bbox_inside_weights_shape = (FreeDimension, self._num_classes * 4)
return [
output_variable(rois_shape,
self.inputs[0].dtype,
self.inputs[0].dynamic_axes,
name="rpn_target_rois_raw",
needs_gradient=False),
output_variable(labels_shape,
self.inputs[0].dtype,
self.inputs[0].dynamic_axes,
name="label_targets_raw",
needs_gradient=False),
output_variable(bbox_targets_shape,
self.inputs[0].dtype,
self.inputs[0].dynamic_axes,
name="bbox_targets_raw",
needs_gradient=False),
output_variable(bbox_inside_weights_shape,
self.inputs[0].dtype,
self.inputs[0].dynamic_axes,
name="bbox_inside_w_raw",
needs_gradient=False)
]
def infer_outputs(self):
return [
C.output_variable(self.inputs[0].shape, self.inputs[0].dtype,
self.inputs[0].dynamic_axes),
C.output_variable(self.inputs[0].shape, self.inputs[0].dtype,
self.inputs[0].dynamic_axes)
]
def infer_outputs(self):
return [
C.output_variable(C.FreeDimension, self.inputs[0].dtype,
self.inputs[0].dynamic_axes),
C.output_variable(C.FreeDimension, self.inputs[0].dtype,
self.inputs[0].dynamic_axes)
]
def infer_outputs(self):
# sampled rois (0, x1, y1, x2, y2)
# for CNTK the proposal shape is [4 x roisPerImage], and mirrored in Python
rois_shape = (cfg["TRAIN"].RPN_POST_NMS_TOP_N, 4)
#rois_shape = (FreeDimension, 4)
# labels
# for CNTK the labels shape is [1 x roisPerImage], and mirrored in Python
labels_shape = (cfg["TRAIN"].RPN_POST_NMS_TOP_N, self._num_classes)
#labels_shape = (FreeDimension, self._num_classes)
# bbox_targets
bbox_targets_shape = (cfg["TRAIN"].RPN_POST_NMS_TOP_N, self._num_classes * 4)
#bbox_targets_shape = (FreeDimension, self._num_classes * 4)
# bbox_inside_weights
bbox_inside_weights_shape = (cfg["TRAIN"].RPN_POST_NMS_TOP_N, self._num_classes * 4)
#bbox_inside_weights_shape = (FreeDimension, self._num_classes * 4)
return [output_variable(rois_shape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="rpn_target_rois_raw", needs_gradient=False),
output_variable(labels_shape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="label_targets_raw", needs_gradient=False),
output_variable(bbox_targets_shape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="bbox_targets_raw", needs_gradient=False),
output_variable(bbox_inside_weights_shape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="bbox_inside_w_raw", needs_gradient=False)]
def infer_outputs(self):
# This is a necessary work around since after cloning the cloned inputs are just place holders without the proper shape
if self._cfm_shape is None:
self._cfm_shape = self.inputs[0].shape
height, width = self._cfm_shape[-2:]
if DEBUG:
print('AnchorTargetLayer: height', height, 'width', width)
A = self._num_anchors
# labels
labelShape = (1, A, height, width)
# Comment: this layer uses encoded labels, while in CNTK we mostly use one hot labels
# bbox_targets
bbox_target_shape = (1, A * 4, height, width)
# bbox_inside_weights
bbox_inside_weights_shape = (1, A * 4, height, width)
return [
output_variable(labelShape,
self.inputs[0].dtype,
self.inputs[0].dynamic_axes,
name="objectness_target",
needs_gradient=False),
output_variable(bbox_target_shape,
self.inputs[0].dtype,
self.inputs[0].dynamic_axes,
name="rpn_bbox_target",
needs_gradient=False),
output_variable(bbox_inside_weights_shape,
self.inputs[0].dtype,
self.inputs[0].dynamic_axes,
name="rpn_bbox_inside_w",
needs_gradient=False),
]
def infer_outputs(self):
height, width = self.inputs[0].shape[-2:]
if DEBUG:
print('AnchorTargetLayer: height', height, 'width', width)
A = self._num_anchors
# labels
labelShape = (1, A, height, width)
# Comment: this layer uses encoded labels, while in CNTK we mostly use one hot labels
# bbox_targets
bbox_target_shape = (1, A * 4, height, width)
# bbox_inside_weights
bbox_inside_weights_shape = (1, A * 4, height, width)
return [
output_variable(labelShape,
self.inputs[0].dtype,
self.inputs[0].dynamic_axes,
name="objectness_target",
needs_gradient=False),
output_variable(bbox_target_shape,
self.inputs[0].dtype,
self.inputs[0].dynamic_axes,
name="rpn_bbox_target",
needs_gradient=False),
output_variable(bbox_inside_weights_shape,
self.inputs[0].dtype,
self.inputs[0].dynamic_axes,
name="rpn_bbox_inside_w",
needs_gradient=False),
]
def infer_outputs(self):
output_vars = [C.output_variable(self.inputs[0].shape, self.inputs[0].dtype, self.inputs[0].dynamic_axes)]
self.action, self.actionArg = self.multiFunc(self.inputs[0])
self.grad, self.gradArg, self.gradRoot = self.gradFunc(self.inputs[0])
return output_vars
def infer_outputs(self):
return [
C.output_variable(self.inputs[0].shape,
self.inputs[0].dtype,
self.inputs[0].dynamic_axes,
needs_gradient=False)
]
def infer_outputs(self):
outputVar = [
C.output_variable(self.inputs[idx].shape,
self.inputs[idx].dtype,
self.inputs[idx].dynamic_axes,
name='outSimpleUdf')
for idx in range(len(self.inputs))
]
return outputVar
def infer_outputs(self):
self.count = self.count + 1
outputVar = [
C.output_variable(self.inputs[1].shape,
self.inputs[1].dtype,
self.inputs[1].dynamic_axes,
name='outDummyLayer')
]
return outputVar
def infer_outputs(self):
# rois blob: holds R regions of interest, each is a 5-tuple
# (n, x1, y1, x2, y2) specifying an image batch index n and a
# rectangle (x1, y1, x2, y2)
# for CNTK the proposal shape is [4 x roisPerImage], and mirrored in Python
proposalShape = (FreeDimension, 4)
return [output_variable(proposalShape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="rpn_rois_raw", needs_gradient=False)]
def infer_outputs(self):
# rois blob: holds R regions of interest, each is a 5-tuple
# (n, x1, y1, x2, y2) specifying an image batch index n and a
# rectangle (x1, y1, x2, y2)
# for CNTK the proposal shape is [4 x roisPerImage], and mirrored in Python
proposalShape = (FreeDimension, 4)
return [output_variable(proposalShape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="rpn_rois_raw", needs_gradient=False)]
def infer_outputs(self):
if not self.shape:
self.shape = self.inputs[0].shape[:-1] + (
self.inputs[1].shape[-1], )
return [
ct.output_variable(self.shape,
np.float32,
self.inputs[1].dynamic_axes,
name=self.name + '_output_shape')
]
def infer_outputs(self):
output_vars = [
C.output_variable(self.inputs[0].shape, self.inputs[0].dtype,
self.inputs[0].dynamic_axes)
]
self.action, self.actionArg = self.multiFunc(self.inputs[0])
self.grad, self.gradArg, self.gradRoot = self.gradFunc(self.inputs[0])
return output_vars
def infer_outputs(self):
height, width = self.inputs[0].shape[-2:]
if DEBUG:
print('AnchorTargetLayer: height', height, 'width', width)
A = self._num_anchors
# labels
labelShape = (1, A, height, width)
# Comment: this layer uses encoded labels, while in CNTK we mostly use one hot labels
# bbox_targets
bbox_target_shape = (1, A * 4, height, width)
# bbox_inside_weights
bbox_inside_weights_shape = (1, A * 4, height, width)
return [output_variable(labelShape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="objectness_target", needs_gradient=False),
output_variable(bbox_target_shape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="rpn_bbox_target", needs_gradient=False),
output_variable(bbox_inside_weights_shape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="rpn_bbox_inside_w", needs_gradient=False),]
def infer_outputs(self):
# rois blob: holds R regions of interest, each is a 5-tuple
# (n, x1, y1, x2, y2) specifying an image batch index n and a
# rectangle (x1, y1, x2, y2)
# for CNTK the proposal shape is [4 x roisPerImage], and mirrored in Python
# cfg_key = str(self.phase) # either 'TRAIN' or 'TEST' --> use FreeDimension and set output size in fwd
proposalShape = (cfg["TRAIN"].RPN_POST_NMS_TOP_N, 4)
#proposalShape = (FreeDimension, 4)
return [output_variable(proposalShape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="rpn_rois_raw", needs_gradient=False)] # , name="rpn_rois" | name="rpn_rois_raw"
def infer_outputs(self):
# This is a necessary work around since anfter cloning the cloned inputs are just place holders without the proper shape
if self._cfm_shape is None:
self._cfm_shape = self.inputs[0].shape
height, width = self._cfm_shape[-2:]
if DEBUG:
print('AnchorTargetLayer: height', height, 'width', width)
A = self._num_anchors
# labels
labelShape = (1, A, height, width)
# Comment: this layer uses encoded labels, while in CNTK we mostly use one hot labels
# bbox_targets
bbox_target_shape = (1, A * 4, height, width)
# bbox_inside_weights
bbox_inside_weights_shape = (1, A * 4, height, width)
return [output_variable(labelShape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="objectness_target", needs_gradient=False),
output_variable(bbox_target_shape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="rpn_bbox_target", needs_gradient=False),
output_variable(bbox_inside_weights_shape, self.inputs[0].dtype, self.inputs[0].dynamic_axes,
name="rpn_bbox_inside_w", needs_gradient=False),]
def infer_outputs(self):
# rois blob: holds R regions of interest, each is a 5-tuple
# (n, x1, y1, x2, y2) specifying an image batch index n and a
# rectangle (x1, y1, x2, y2)
# for CNTK the proposal shape is [4 x roisPerImage], and mirrored in Python
# cfg_key = str(self.phase) # either 'TRAIN' or 'TEST' --> use FreeDimension and set output size in fwd
proposalShape = (cfg["TRAIN"].RPN_POST_NMS_TOP_N, 4)
#proposalShape = (FreeDimension, 4)
return [
output_variable(proposalShape,
self.inputs[0].dtype,
self.inputs[0].dynamic_axes,
name="rpn_rois_raw",
needs_gradient=False)
] # , name="rpn_rois" | name="rpn_rois_raw"
def infer_outputs(self):
impl_func_output = self.impl_func.output
return [C.output_variable(impl_func_output.shape, impl_func_output.dtype, impl_func_output.dynamic_axes)]
def infer_outputs(self):
impl_func_output = self.impl_func.output
return [
C.output_variable(impl_func_output.shape, impl_func_output.dtype,
impl_func_output.dynamic_axes)
]
def infer_outputs(self):
conversation_batch_axis = C.Axis.default_batch_axis()
return [
C.output_variable((C.FreeDimension, ) + self.inputs[0].shape,
self.inputs[0].dtype, [conversation_batch_axis])
]
def infer_outputs(self):
return [C.output_variable(C.FreeDimension, self.inputs[0].dtype, self.inputs[0].dynamic_axes),
C.output_variable(C.FreeDimension, self.inputs[0].dtype, self.inputs[0].dynamic_axes)]
def infer_outputs(self):
return [output_variable(self.inputs[0].shape, self.inputs[0].dtype, self.inputs[0].dynamic_axes, name='rpn_obj_prob'),
output_variable(self.inputs[1].shape, self.inputs[1].dtype, self.inputs[1].dynamic_axes, name='rpn_obj_targets', needs_gradient=False)]
def infer_outputs(self):
return [
output_variable((), self.inputs[0].dtype,
self.inputs[0].dynamic_axes)
]
def infer_outputs(self):
self.count = self.count + 1
outputVar = [C.output_variable(self.inputs[1].shape, self.inputs[1].dtype,
self.inputs[1].dynamic_axes, name='outDummyLayer')]
return outputVar
def infer_outputs(self):
conversation_batch_axis = C.Axis.default_batch_axis()
return [C.output_variable((C.FreeDimension,) + self.inputs[0].shape, self.inputs[0].dtype, [conversation_batch_axis])]
def infer_outputs(self):
return [C.output_variable(self.inputs[0].shape, self.inputs[0].dtype, self.inputs[0].dynamic_axes, needs_gradient=False)]
def infer_outputs(self):
outputVar = [C.output_variable(self.inputs[idx].shape, self.inputs[idx].dtype,
self.inputs[idx].dynamic_axes, name='outDummyLayer') for idx in range(len(self.inputs))]
return outputVar
def infer_outputs(self):
return [output_variable(self.inputs[0].shape, self.inputs[0].dtype, self.inputs[0].dynamic_axes)]
