mingjie.xiao

I think I've found your strategy in [generate_data.py](https://github.com/nywang16/Pixel2Mesh/blob/master/data_generation/generate_data.py). But I still wonder that why you use different point numbers to sample a mesh instead of a fixed number like 16384...

sry.. I use my own dataset that renders images and samples pointclouds from Shapenet Dataset. I suggest that u can dig into Shapenet dataset and see how to deal with...

@BGHB You may follow [PSGN](https://github.com/fanhqme/PointSetGeneration) to compile it from .cpp

HI @luotuoqingshan , As far as I know, training data of PSGN is much more larger than Shapenet dataset. Jack renders and samples each single model of Shapenet dataset using...

@luotuoqingshan , I retrained PSGN with my own data. But due to the differences of data distribution and normalization, I didn't get the same CD numbers or EMD numbers as...

@luotuoqingshan , I regard that parameters as hyper-parameters. As far as I know, only PSGN changes the hyper-parameter between forward and backward. Most works directly define the loss as the...

@luotuoqingshan I can only come up with two thoughts. 1. I forgot to normalize my input images to 0.~1. when I was doing predicting. 2. Maybe your network had not...

@luotuoqingshan The speed depends on your training data if you did not modify PSGN. The official implementation works for me. And if you did some modification on PSGN then I...

I know that cd is a squraed distance while emd is a actual distance. I compute emd following implementation [here](https://github.com/Vahe1994/latent_3d_points/blob/master/external/structural_losses/tf_approxmatch.cpp). with the following code. `match = approx_match(xyz1, xyz2)` `emd =...