Random crop

This block randomly crops input images to a given height and width. Randomly cropping images is an efficient and low-cost way to perform data augmentation, which can increase accuracy.

Using the Random crop block adds more variation to the training dataset, and is helpful for both large and small datasets. T he random location of cropping on the image is chosen per batch, so all images within one batch will be cropped in the same way. Note that this block is only activated during training and will be disabled during testing and deployment.

Parameters

Height: Output height of the cropped image. Must be equal to or smaller than the original images.

Weight: Output width of the cropped image. Must be equal to or smaller than the original images.

Was this page helpful?
YesNo