Function Differences with tf.image.resize_images

tf.image.resize_images

tf.image.resize_images(
    images,
    size,
    method=ResizeMethodV1.BILINEAR,
    align_corners=False,
    preserve_aspect_ratio=False,
    name=None
)

For more information, see tf.image.resize_images.

mindspore.dataset.vision.c_transforms.Resize

class mindspore.dataset.vision.c_transforms.Resize(
    size,
    interpolation=Inter.LINEAR
)

For more information, see mindspore.dataset.vision.c_transforms.Resize.

Differences

TensorFlow: Resize the image to the specified size. It supports aligning the centers of the 4 corner pixels and preserving the aspect ratio.

MindSpore: Resize the image to the specified size. It will keep the aspect ratio when the input size is a single integer.

Code Example

# The following implements Resize with MindSpore.
import numpy as np
import mindspore.dataset as ds
from mindspore.dataset.vision import Inter

image = np.random.random((28, 28, 3))
result = ds.vision.c_transforms.Resize((14, 14), Inter.BICUBIC)(image)
print(result.shape)
# (14, 14, 3)

# The following implements resize_images with TensorFlow.
import tensorflow as tf
from tensorflow.image import ResizeMethod

image = tf.random.normal((28, 28, 3))
result = tf.image.resize_images(image, (14, 14), ResizeMethod.BICUBIC)
print(result.shape)
# (14, 14, 3)