torchvision_sunner.transforms.function¶

This pages shows the parameters of the function which torchvision_sunner provide. Unlike the augmentation class, these function call be call directly. Also, you just need to import for single module:

import torchvision_sunner.transforms as sunnertransforms

torchvision_sunner.transforms.function.asImg [source]¶

This function can transfer the computed tensor into numpy image directly, and resize the image into ideal size. This function may be appropriatly used after the final result of network. You should notice the value range should locate in [-1, 1]. Besides, the rank format of input tensor should be BCHW. Here is the usage:

import cv2
rendered_img = asImg(net_output_img, [320, 640])
cv2.imshow('show_img_window', rendered_img)
cv2.waitKey()

Parameters¶

tensor (np.ndarray or torch.Tensor) - The tensor you want to deal with.
size (tuple or list) - The size of output image

torchvision_sunner.utils.quiet [source]¶

This function can mute all information preview in the terminal. Here is the usage:

sunnertransforms.quiet()

torchvision_sunner.transforms.function.show [source]¶

This function utilize the previous function, and show the image directly. This function may be appropriatly used after the final result of network. You should notice the value range should locate in [-1, 1]. Besides, the rank format of input tensor should be BCHW. Here is the usage:

# Do the forwarding first
img = net(data)

# Show it in 2 * 4 grid
show(img[:8], row=2, column=4)

Parameters¶

tensor (np.ndarray or torch.Tensor) - The tensor you want to deal with.
row (Int) - The number of row you want to extend
column (Int) - The number of column you want to extend
title (Int) - The title string in the demonstration image
sec (Int) - The time you want to pause. Set -1 if you want to wait until press any key