How to find pixel coordinate to world coordinate?

[tgrmev]

Hello, I come across with your repository. I saw you explained camera calibration inside camera projection file. I want to do inverse process of what you did. I need to get world coordinate from pixel coordinate. Unfortunately, I could not find the correct locations. I want to ask you that, you add additional element of world coordinate matrix to multiple with the 3x4 tranformation matrix, to implement what I want, still do I need to use 3x4 transformation matrix or not? For my case, I will multiply inverse transformation matrix with 3x1 pixel matrix. If I need to use 3x4 transformation matrix like you did, how can I take the inverse of that matrix? With np.linalg.pinv or do I need to add additional line like (0,0,0,1) along axis 0 then take the inverse with np.linalg.inv? I hope my question is clear for you. If I could not explaine clearly, sorry for that.