Python roi_pool_grad Examples

Example #1

File: roi_pooling_op_grad.py Project: thunderock/FRCNN_git

def _roi_pool_grad(op, grad, _):
    #The gradients for `roi_pool`.
    #Args:
    #  op: The `roi_pool` `Operation` that we are differentiating, which we can use
    #    to find the inputs and outputs of the original op.
    #  grad: Gradient with respect to the output of the `roi_pool` op.
    #Returns:
    #  Gradients with respect to the input of `zero_out`.

    data = op.inputs[0]
    rois = op.inputs[1]
    orientations = op.inputs[2]
    argmax = op.outputs[1]
    pooled_height = op.get_attr('pooled_height')
    pooled_width = op.get_attr('pooled_width')
    spatial_scale = op.get_attr('spatial_scale')

    # compute gradient
    data_grad = roi_pooling_op.roi_pool_grad(data, rois, argmax, grad,
                                             orientations, pooled_height,
                                             pooled_width, spatial_scale)

    # data_grad contains the gradients with respect to the VGG output and the two 'none's indicate that there is no
    # gradient with respect to the bounding box positions or rotations
    return [data_grad, None,
            None]  # List of one Tensor, since we have one input

Example #2

File: roi_pooling_op_grad.py Project: ChuanqiL/TFFRCNN

def _roi_pool_grad(op, grad, _):
  """The gradients for `roi_pool`.
  Args:
    op: The `roi_pool` `Operation` that we are differentiating, which we can use
      to find the inputs and outputs of the original op.
    grad: Gradient with respect to the output of the `roi_pool` op.
  Returns:
    Gradients with respect to the input of `zero_out`.
  """
  data = op.inputs[0]
  rois = op.inputs[1]
  argmax = op.outputs[1]
  pooled_height = op.get_attr('pooled_height')
  pooled_width = op.get_attr('pooled_width')
  spatial_scale = op.get_attr('spatial_scale')

  # compute gradient
  data_grad = roi_pooling_op.roi_pool_grad(data, rois, argmax, grad, pooled_height, pooled_width, spatial_scale)

  return [data_grad, None]  # List of one Tensor, since we have one input

Example #3

File: roi_pooling_op_grad.py Project: alicezywang/cognition

def _roi_pool_grad(op, grad, _):
  """The gradients for `roi_pool`.
  Args:
    op: The `roi_pool` `Operation` that we are differentiating, which we can use
      to find the inputs and outputs of the original op.
    grad: Gradient with respect to the output of the `roi_pool` op.
  Returns:
    Gradients with respect to the input of `zero_out`.
  """
  data = op.inputs[0]
  rois = op.inputs[1]
  argmax = op.outputs[1]
  pooled_height = op.get_attr('pooled_height')
  pooled_width = op.get_attr('pooled_width')
  spatial_scale = op.get_attr('spatial_scale')

  # compute gradient
  data_grad = roi_pooling_op.roi_pool_grad(data, rois, argmax, grad, pooled_height, pooled_width, spatial_scale)

  return [data_grad, None]  # List of one Tensor, since we have one input