tiling | Tiling provides a simple window tiling system for linux

If you duplicate 12 times, you'll have 13 images in toto. Try -duplicate 11 instead.

The themed_title_bar setting controls whether the theme you're using has the ability to change the color of the title bar or not to match the theme (for themes that support it). On Windows this also has the effect of hiding the menu and placing it in a rolled up "hamburger menu" in the title bar.

On Linux, there is also support for custom title bars irrespective of that which only enables itself if it detects that you're using an appropriate window manager. In cases where it detects incorrectly (you want the feature but it's not enabled) or you want to remove it because you don't want it (which is sounds like you're running into here), you need to adjust the value of this setting:

Surface.blit wants a rect as area argument. You pass a string.

What you could do is create a map to translate the strings to the right rects that you already defined, something like this:

I replaced the gradient estimation code with my own solution gradient and the code works now. Calculating errors can be tricky. As on can see on the histogram (note the log-scale) at the bottom, for most pixels the relative error is smaller than 10^-4 but where the gradient is close to zero, the relative error explodes. The problem with max(rel_err) and mean(rel_err) is, that they are both easily perturbed by outlier pixels. Better measures for if the order of magnitude is most relevant are the geometric mean and median over all non-zero pixels.

In second pass, depth attachment description, it "loads" image in undefined state:

Content-Type was incorrectly set. I was using Postman, latest update seems to have broken it. It was working fine before. Tried with curl and it worked

I'm posting this as an answer because it's long, but it's not literally an answer. I'm hoping this will lead you to the next step in your design process.

Here are the design decisions you face, It's clear from your code that you can generate a list of triangle coordinates. Good, what next? You probably know the bounding box of your triangles (largest w and h) advance, so you can certainly create a set of images that contain your triangles, masked off with a black background or alpha=0 background. You could just copy the bounding rectangle to an image, then create a mask using the triangle as a path, and set the alpha to 0 outside of the mask. opencv should be able to do that.

But after you have those, what then? You talked about matching the edges. That's complicated. I suppose you could extract a vector of pixels from the three edges of each triangle, and then do some kind of approximate comparison.

If you do find matches that allow you to stitch together a composite, it is possible (assuming you have alpha=0 in the backgrounds) to blit all of these triangles back into some larger image, kind of like quilting. openvc can do block copy with alpha blending.

So, in the end, I think your problem is achievable, but it's going to be a lot of work, and probably more than we can offer here.

Try a value slightly bigger than 1px

When starting the three windows, use xterm's -class option to modify the WM_CLASS property it attaches to its window, choosing a different and unique class for each of the three. Then you can select for that class in your manage hook to only match the right window.