In this setting, the results are not deterministic. Do you have some idea on where does the randomness come from? And is there some way to avoid such randomness?
Below is an example code snippet to see the problem. When you run it multiple times, you will get inconsistent results. (Here I hardcode the camera pose and light pose here to avoid some additional library to calculate it, but you can use any pose)
In this setting, the results are not deterministic. Do you have some idea on where does the randomness come from? And is there some way to avoid such randomness?
Below is an example code snippet to see the problem. When you run it multiple times, you will get inconsistent results. (Here I hardcode the camera pose and light pose here to avoid some additional library to calculate it, but you can use any pose)
You can use this model from ShapeNet to test https://drive.google.com/file/d/1M0ncX1IXSxcpGijlDm58MCr36CaGts2S/view?usp=sharing
Here are the two inconsistent results: