antao97 / UnsupervisedPointCloudReconstruction

Experiments on unsupervised point cloud reconstruction.
MIT License
165 stars 27 forks source link

Any advice for adjustments to the folding parameters when increasing the number of points? #5

Closed ninasinger closed 2 years ago

ninasinger commented 2 years ago

I'm trying to reconstruct 16384 points. Right now the points are looking very "artificial", clear lines of points almost like scan lines. Any suggestions on what parameters to update?

antao97 commented 2 years ago

I suggest using sphere as the reconstruction source and using more than three layers.

ninasinger commented 2 years ago

Thanks so much for the response! When you say additional layers, are you referring to an additional downsampling layer and then an additional folding operation or an additional hidden layer to the MLP which is performing the folding operation.

antao97 commented 2 years ago

I think an additional folding layer may help. But I do not fully understand your case. I just think making the network more complex may work.