Class RandomSelectSubpolicy
Defined in File vision.h
Inheritance Relationships
Base Type
public mindspore::dataset::TensorTransform
(Class TensorTransform)
Class Documentation
-
class RandomSelectSubpolicy : public mindspore::dataset::TensorTransform
Choose a random sub-policy from a list to be applied on the input image. A sub-policy is a list of tuples (operation, prob), where operation is a TensorTransform operation and prob is the probability that this operation will be applied. Once a sub-policy is selected, each operation within the sub-policy with be applied in sequence according to its probability.
Public Functions
-
explicit RandomSelectSubpolicy(const std::vector<std::vector<std::pair<TensorTransform*, double>>> &policy)
Constructor.
- Parameters
policy – [in] Vector of sub-policies to choose from, in which the TensorTransform objects are raw pointers.
样例/* Define operations */ auto invert_op(new vision::Invert()); auto equalize_op(new vision::Equalize()); std::vector<std::pair<TensorTransform *, double>> policy = {{invert_op, 0.5}, {equalize_op, 0.4}}; vision::RandomSelectSubpolicy random_select_subpolicy_op = vision::RandomSelectSubpolicy({policy}); /* dataset is an instance of Dataset object */ dataset = dataset->Map({random_select_subpolicy_op}, // operations {"image"}); // input columns
Constructor.
- Parameters
policy – [in] Vector of sub-policies to choose from, in which the TensorTransform objects are shared pointers.
样例/* Define operations */ std::shared_ptr<TensorTransform> invert_op(new vision::Invert()); std::shared_ptr<TensorTransform> equalize_op(new vision::Equalize()); std::shared_ptr<TensorTransform> resize_op(new vision::Resize({15, 15})); auto random_select_subpolicy_op = vision::RandomSelectSubpolicy({ {{invert_op, 0.5}, {equalize_op, 0.4}}, {{resize_op, 0.1}} }); /* dataset is an instance of Dataset object */ dataset = dataset->Map({random_select_subpolicy_op}, // operations {"image"}); // input columns
-
explicit RandomSelectSubpolicy(const std::vector<std::vector<std::pair<std::reference_wrapper<TensorTransform>, double>>> &policy)
Constructor.
- Parameters
policy – [in] Vector of sub-policies to choose from, in which the TensorTransform objects are object pointers.
样例/* Define operations */ vision::Invert invert_op = vision::Invert(); vision::Equalize equalize_op = vision::Equalize(); vision::Resize resize_op = vision::Resize({15, 15}); auto random_select_subpolicy_op = vision::RandomSelectSubpolicy({ {{invert_op, 0.5}, {equalize_op, 0.4}}, {{resize_op, 0.1}} }); /* dataset is an instance of Dataset object */ dataset = dataset->Map({random_select_subpolicy_op}, // operations {"image"}); // input columns
-
~RandomSelectSubpolicy() = default
Destructor.
-
explicit RandomSelectSubpolicy(const std::vector<std::vector<std::pair<TensorTransform*, double>>> &policy)