Lab to RGB conversion problem

[mycarta]

Hi Greg

When I run the RGB to Lab conversion below (it is the same as the one I showed you in my previous request, but this time on a 256x3 array), it works wonderfully.

LAB=[]
for rgb_val in RGB255:
    rgb_r, rgb_g, rgb_b = rgb_val
    rgb_color = RGBColor(rgb_r, rgb_g, rgb_b)
    lab_color = rgb_color.convert_to('lab')
    LAB.append(lab_color.get_value_tuple())

Now I am trying to convert a 256x3 Lab array to RGB. This is my Lab array (essentially a sigmoid function going from black to white):

x = np.linspace(0,10,256)
L=1/(1+np.exp(-(x-5)/1))
a=np.zeros(256)
b=np.zeros(256)

I use this code below.

from colormath.color_objects import LabColor
from colormath.color_objects import RGBColor
RGB=[]
for Lab_val in Lab:
    Lab_l, Lab_a, Lab_b = Lab_val
    lab = LabColor(Lab_l, Lab_a, Lab_b)
    rgb=lab_color.convert_to('rgb')
    RGB.append(rgb.get_value_tuple())

As far I can tell this snippet should work, but when I look at the output RGB, it is made of all zeros.
I use colormath in Enthought Canopy so I am still running version 1.0.8
Do you see anything I am missing?

Thanks,
Matteo

[gtaylor]

The RGB conversions leave a lot to be desired in 1.x. I'd suggest giving 2.0.0 a try. You'll have to make some changes, but they aren't extreme. Here's what conversions look like now:

http://python-colormath.readthedocs.org/en/latest/conversions.html#example

I won't be maintaining the 1.x line, so I'll close this issue. If you try again with 2.x and still run into this, re-open with what you've got. At that point, we'd need to start looking into cranking up the logger to the DEBUG level and see where the conversion is diverging from expectation.

[mycarta]

OK thanks
I will have to wait until 2.x is populated to Canopy. I'll keep you in the loop.

[gtaylor]

Do you know if they need package maintainers to tell them to pull updates? I'd definitely prefer to scrub all traces of colormath 1.x, it needs to go away.

[mycarta]

This is what I see on their package manager:

However, the PyPi mirror does show 2.0.0:

I sent them a support request, let's see what they say.

[gtaylor]

OK, let me know if I need to get in touch with them somehow. Feel free to reference this issue. The 1.x series will not see maintenance, and has some serious issues in RGB land that 2.x solves.

[mycarta]

OK thanks. They do say the community packages come as is. Hopefully the 2.0.0 will show up soon (perhaps they only pull from the mirror every so often, or once every update). As for me, if it does not happen soon, I'll have to try and install manually.

[mycarta]

Hi Greg

Below is the response I got from Enthought. Not the greatest. It looks like 1.x could sit around in Canopy for quite a while.
I'm going to take my chances with the manual installation, but if you want you can ask people in your social media network that may be paying subscribers of Canopy, to put in a request out update colormath.
Cheers

Hi Matteo,

These 11,000+ packages are considered an incidental premium for subscribers, and as such are a much lower priority than all other build tasks. The build team update the list irregularly, but typically once or twice a year. A much higher priority for the build team is to make infrastructure improvements which will make it possible to update this and (more importantly) the main repository, much more frequency.

If many subscribers ask about a particular package, then we may give that package special treatment; you are the only person who has asked about ColorMath.

[gtaylor]

Wow, that's pretty bad. Is this a popular service? I guess I'm not too sure what this gets you versus a local ipython install.

[mycarta]

I like working in Canopy. it's awesome for the kind of things many geoscientist use programming for.
https://www.enthought.com/products/canopy/
Unfortunately for me it is also the first and only way I learned to use Python to begin with. But this may push me to learn to work in both modes.

[gtaylor]

No idea what to tell you, there. But if you like the visual aspect of it, ipython is worth checking out: http://ipython.org/