Reduce color palette converter install#

Note _to_xyz assumes floating point input in the range of so additional pre-processing is needed: # convert to float32 The following is the conversion from RGB to CIE XYZ through _to_xyz. Some image software and camera manufacturors might prefer BGR, which can be obtained through _to_bgr: bgr = _to_bgr(image)ĬIE XYZ (or CIE 1931 XYZ is a common color space used in many image processing programs. Plt.imshow(tf.squeeze(grayscale, axis=-1), cmap='gray') # use tf.squeeze to remove last channel for plt.imshow to display: The image can be displayed by: import matplotlib.pyplot as pltĪn RGB image can be converted to Grayscale to reduce the channel from 3 to 1 with _to_grayscale: grayscale = _to_grayscale(image) Image = tf.code_jpeg(tf.io.read_file('sample.jpg')) Read and decode the image into a uint8 Tensor of shape (213, 320, 3) import tensorflow as tf The following will download the image and save to local disk as sample.jpg: curl -o sample.jpg -L ls -ls sample.jpg

The image example used in this tutorial is a cat in the snow, though it could be replaced by any JPEG images.

Reduce color palette converter install#

Setup Install required Packages, and restart runtime pip install -q tensorflow-io Download the sample image The tensorflow-io package provides a list of color space conversions APIs that can be used to prepare and augment the image data. While RGB is the most common color space, in manay situations the model performs better when switching to alternative color spaces such as YUV, YCbCr, XYZ (CIE), etc. In computer vision, the selected color space could have a significant the performance of the model.