mindspore.dataset.vision.BoundingBoxAugment

View Source On Gitee
class mindspore.dataset.vision.BoundingBoxAugment(transform, ratio=0.3)[source]

Apply a given image processing operation on a random selection of bounding box regions of a given image.

Parameters

  • transform (TensorOperation) – Transformation operation to be applied on random selection of bounding box regions of a given image.

  • ratio (float, optional) – Ratio of bounding boxes to apply augmentation on. Range: [0.0, 1.0]. Default: 0.3.

Raises

  • TypeError – If transform is an image processing operation in mindspore.dataset.vision .

  • TypeError – If ratio is not of type float.

  • ValueError – If ratio is not in range [0.0, 1.0].

  • RuntimeError – If given bounding box is invalid.

Supported Platforms:

CPU

Examples

>>> import numpy as np
>>> import mindspore.dataset as ds
>>> import mindspore.dataset.vision as vision
>>>
>>> # Use the transform in dataset pipeline mode
>>> data = np.random.randint(0, 255, size=(100, 100, 3)).astype(np.float32)
>>> numpy_slices_dataset = ds.NumpySlicesDataset(data, ["image"])
>>> func = lambda img: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(np.float32))
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[func],
...                                                 input_columns=["image"],
...                                                 output_columns=["image", "bbox"])
>>> # set bounding box operation with ratio of 1 to apply rotation on all bounding boxes
>>> bbox_aug_op = vision.BoundingBoxAugment(vision.RandomRotation(90), 1)
>>> # map to apply ops
>>> numpy_slices_dataset = numpy_slices_dataset.map(operations=[bbox_aug_op],
...                                                 input_columns=["image", "bbox"],
...                                                 output_columns=["image", "bbox"])
>>> for item in numpy_slices_dataset.create_dict_iterator(num_epochs=1, output_numpy=True):
...     print(item["image"].shape, item["image"].dtype)
...     print(item["bbox"].shape, item["bbox"].dtype)
...     break
(100, 100, 3) float32
(1, 4) float32
>>>
>>> # Use the transform in eager mode
>>> data = np.array([[0, 1, 2, 3, 4, 5], [0, 1, 2, 3, 4, 5]], dtype=np.uint8).reshape((3, 4))
>>> data = data.astype(np.float32)
>>> func = lambda img, bboxes: (data, np.array([[0, 0, data.shape[1], data.shape[0]]]).astype(bboxes.dtype))
>>> func_data, func_bboxes = func(data, data)
>>> output = vision.BoundingBoxAugment(transforms.Fill(100), 1.0)(func_data, func_bboxes)
>>> print(output[0].shape, output[0].dtype)
(3, 4) float32
>>> print(output[1].shape, output[1].dtype)
(1, 4) float32
Tutorial Examples: