mindspore.dataset.vision.Decode

class mindspore.dataset.vision.Decode(to_pil=False)[source]

Decode the input image in RGB mode. Supported image formats: JPEG, BMP, PNG, TIFF, GIF(need to_pil=True ), WEBP(need to_pil=True ).

Parameters

to_pil (bool, optional) – decode to PIL Image. Default: False.

Raises

RuntimeError – If given tensor is not a 1D sequence.
RuntimeError – If the input is not raw image bytes.
RuntimeError – If the input image is already decoded.

Supported Platforms:: CPU

Examples

>>> # Eager usage
>>> import numpy as np
>>> raw_image = np.fromfile("/path/to/image/file", np.uint8)
>>> decoded_image = vision.Decode()(raw_image)
>>>
>>> # Pipeline usage
>>> transforms_list = [vision.Decode(), vision.RandomHorizontalFlip()]
>>> image_folder_dataset = image_folder_dataset.map(operations=transforms_list,
...                                                 input_columns=["image"])