Open LeroyChou opened 4 months ago
Hi there The normalized coordinates serve as input to the key model. Normalizing ensures that the scale of the inputs is independent of the scale of the CAD model.
Given that each CAD model utilizes a separate MLP, would it be neccessary to make sure that the input is independent of the scale?
Scaling of inputs to neural networks can matter. But yes, it might still work :)
Hello, I want to express my gratitude for your excellent work.
I recently observed that you normalize the objects' coordinates using offset and scale in this line. Why not use the original coordinates? I would greatly appreciate it if you could provide the rationale behind this decision. Thank you very much.