This is how it looks 🙂

Haha, ok, DeepSource is a little picky about the white space in the docstring. I'll fix that. I also forgot to add this to the documentation

@HaleySchuhl I'm missing an example image for the documentation. I already prepared everything in the md file, but I could not find the original image that was used for the visualize.pseudocolor() doc page.

I'm calling "img/documentation_images/pseudocolor/pseudo_on_rgb_image.jpg" as image, but this image does not exist yet.

edited:
Thanks @rayn-alex 👏 the example image here looks awesome! I found the original images used in that doc page but there's not an RGB version of that Top View plant, not with the same field of view anyways. There's no harm in using a different example for the RGB use-case in these docs, and I think you're example image above is great if you'd be willing to use that.

We also discussed this morning that adding a new parameter bg_image is technically a breaking change, which we are trying to avoid between major version releases. Instead, it might be most intuitive to update the existing background parameter to accept either `"white", "black", "image", OR an array object. Then if the array object is provided, it'll be used for background. Thoughts?

This absolutely makes sense!
I'll make the edits.

@HaleySchuhl
I just started working on this and realized that introducing this in the background variable requires substancial changes in the code.

The whole magic of this part needs to be replaced, because background is not always a string anymore.

Is this fine with you?

# Set the background color or type
    bkgd = {
        "BLACK": {
            "image": np.zeros(np.shape(gray_img1), dtype=np.uint8),
            "cmap": "gray"
            },
        "WHITE": {
            "image": np.zeros(np.shape(gray_img1), dtype=np.uint8) + 255,
            "cmap": "gray_r"
            },
        "IMAGE": {
            "image": bg_image1 if bg_image1 is not None else gray_img1,
            "cmap": None if bg_image1 is not None else "gray"
            }
    }
    bkg_img = bkgd[background.upper()]["image"]
    bkg_cmap = bkgd[background.upper()]["cmap"]

Hey @HaleySchuhl
I just moved the whole background checking block a bit lower, and then things did not look as horrible as I expected.

Just one thing:
Should we check for the input image dimensions in this part of the code? There will be an error if the gray x/y shape does not fit to the background array x/y from matplotlib. The question is if we should turn it into something more meaningful for the user.