Gotcha! CV2 and Pyplot Channel Order

While working on the first part of my style-transfer project, I used Open CV’s cv2.imwrite to save images to disk. However, the images seemed to have a weird tint to them. In the figure below, I show the original image (left) alongside the result of loading the image using cv2 and displaying it with pyplot (right):

After reading the documentation for imwrite more carefully, I realised that cv2 was interpreting the channel order as BGR instead of RGB which was what pyplot was expecting. Converting between the two was fairly straightforward:

image = cv2.imread('dummy_image.jpg')
rgb_image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

Then when working with pyplot, you can use rgb_image.

And here is the result:

Note that if you had an image in the RGB space and wanted to save it to disk using cv2, you’d need to convert it to BGR first as cv2 saves images in that format. In general, whether displaying images in a notebook or saving them to disk, if the colour scheme seems off, check the channel order.

If you’d like to play around more, here’s a notebook for you to tinker with!

• 2020 3
• 2019 3
• 2018 11