Video transformations for deep learning

• Works with both image and feature map
• Video input is simply a numpy array (no need of PIL).
• Supports both channels first and channels last

Prerequisites

numpy >= 1.12
cv2

Options

Options	Finished
Read
Compose
RandomCrop
RandomCornerCrop
CenterCrop
FiveCrop
TenCrop
Resize
RandomHorizontalFlip
RandomVerticalFlip
Montage
ToTensor
ToBatchTensor
Normalize