Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Mistakes in the docstrings of cam2pixel and pixel2cam

See original GitHub issue

Hey everyone. I was going through the codebase and I noticed a few mistakes in the docstrings of the functions cam2pixel and pixel2cam which are available in kornia/geometry/camera/pinhole.py.

The function cam2pixel mentions the following about the return value in its docstring: https://github.com/kornia/kornia/blob/9d594687db704845c3668d76b6e481666277e375/kornia/geometry/camera/pinhole.py#L582-L584

But on reading the code of the function cam2pixel, it looks like the output is in pixel coordinates. The output can exceed the range [-1, 1]. The function cam2pixel is used in kornia/geometry/depth.py. Here, we can see that the range normalization is applied to pixel_coords_src, which is the output of cam2pixel.

https://github.com/kornia/kornia/blob/9d594687db704845c3668d76b6e481666277e375/kornia/geometry/depth.py#L304-L310

Also, the function pixel2cam mentions the following about the return value in its docstring: https://github.com/kornia/kornia/blob/9d594687db704845c3668d76b6e481666277e375/kornia/geometry/camera/pinhole.py#L557-L559

But on reading the code of the function pixel2cam, it seems like the output is in XYZ camera coordinates. The docstring mentioned that the output would be a tensor of (u, v, 1) cam coordinates which does not make sense. The output seems to be (x, y, z) cam coordinates.

In summary, I think the return value descriptions in the docstrings of cam2pixel and pixel2cam are incorrect. I have not tried running the code though, I’m just mentioning my observations based on my reading. Do let me know your thoughts! I would be happy to send in a PR to fix this if the issue is verified.