Add RandomResizedCrop

[innat]

keras-team/keras-cv#131

[amitsrivastava78]

import keras

FIXED_CROP_SIZE = (180, 180)
TARGET_SIZE = (224, 224)

model = keras.Sequential([
keras.Input(shape=(None, None, 3)),
# 1. Randomly crop a fixed-size window (180x180)
keras.layers.RandomCrop(
height=FIXED_CROP_SIZE[0],
width=FIXED_CROP_SIZE[1],
# Note: Input must be large enough, or use target_height parameter
),
# 2. Resize the result to the final target size
keras.layers.Resizing(
height=TARGET_SIZE[0],
width=TARGET_SIZE[1]
)
])

Check if above helps

[innat]

@amitsrivastava78
Thanks for checking.

They look similar but the way the function works underneath is different. Some discussion are made that mentioned ticket keras-team/keras-cv#131 (comment) - which lead to have this feature in keras-cv.

Here is the torch API.