Need help with pytorch-colors?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

164 Stars 13 Forks MIT License 5 Commits 2 Opened issues


Small utility for color space conversions of PyTorch Tensors and Variables

Services available


Need anything else?

Contributors list

No Data

PyTorch Colors

PyTorch Colors is a simple utility to convert Tensors and Variables between color spaces. Currently all the operations are non-differentiable. PyTorch Colors is in early alpha, and supports the following color spaces: 1. RGB 2. HSV 3. CIE*Lab 4. YUV 5. YCbCr 6. XYZ 7. HED

Please request additional color spaces in the issues or provide a pull request with the implementation. Differentiable versions for each color space conversion will also be added in time.

Installation and requirements

To install PyTorch Colors use the command

python install
inside the repository directory. PyTorch Colors requires the default PyTorch (version > 0.3) installation and scikit-image (version > 0.1) to work.


All PyTorch-Colors methods from color space C1 to color space C2 are named

. Either c1 or c2 must be RGB, so all color spaces can be converted between each other by going through the RGB color space.

All the methods expect a FloatTensor and output a FloatTensor.

 import pytorch_colors as colors

img = torch.randn(1, 3, 256, 256) # example image in RGB img_hsv = colors.rgb_to_hsv(img)

hsv -> ycbcr

img = colors.hsv_to_rgb(img_hsv) img_ycbcr = colors.rgb_to_ycbcr(img)

The supported inputs are either 3 dimensional

(channel, width, height)
or 4 dimensional
(batch_size, channel, width, height)
images. So the following is also valid:
 img = torch.randn(3, 256, 256) # example image in RGB
 img_hsv = colors.rgb_to_hsv(img) # output is also 3 dimensional

The supported inputs are Tensors and Variables, with or without cuda activated and the output should be consistent with the input.

 img = torch.randn(1, 3, 256, 256) # example image in RGB
 v_img = Variable(torch.randn(1, 3, 256, 256)) # example Variable image in RGB
 c_v_img = v_img.cuda() # with CUDA activated

img_hsv = colors.rgb_to_hsv(img) # output is Tensor v_img_hsv = colors.rgb_to_hsv(img) # output is Variable c_v_img_hsv = colors.rgb_to_hsv(img) # output is CUDA Variable

A method with string input is also provided, and can be used for example with script arguments. The string input to convert from c1 to c2 should be

. I.e:
to convert from RGB to HSV.
 import pytorch_colors as colors

parser = argparse.ArgumentParser() parser.add_argument('-c', '--color', help='...') (...) cfg = parser.parse_args() input_ = colors.convert(input_, cfg.color)

You can find more examples in the tests/ folder.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.