Li Ruihui
Li Ruihui
I think it may be about the dataset path.
The only processing is on the training point set (normalizing the points to fit a unit ball), which is a common operation in most point cloud-based analysis. No other processing...
Pls check here. You will find how to work as I described. :) https://github.com/yulequan/PU-Net/tree/master/prepare_data (Mesh Segmentation)
Yes for all :)
You can change the seed number and size of the patch in the Mesh Segmentation. (It is a .sln project). The code is on the main cpp file.
Yes. If possible, pls read the code and test it several times. It should be easier for your understanding. :)
It is in the uniform_loss for triaing and evaluation code for testing
Pls follow the evaluation step. first compile the evluation.cpp, evaluate on the model and then using the evluation.py For more information, you can check the cpp code
patch number (10) is equal to the seed points number and the radius represents the size of each patch. I used both vs2017 and 2019, and it can be compiled...
Thanks for pointing out this issue. After several checking, we found that the per-point discriminator loss tends to result in unstable training. Without it, the performance and result are actually...