search

Colin97 / MSN-Point-Cloud-Completion

Morphing and Sampling Network for Dense Point Cloud Completion (AAAI2020)
Apache License 2.0
417 stars 57 forks source link

Transfer Learning #10

Closed mactavish10 closed 4 years ago

mactavish10 commented 4 years ago

Hello @Colin97 , I was wondering if there's a way to perform transfer learning on the model you provided? Also, could you elaborate on the data prep part? I'm using ply files of my own, not from ShapeNet.

Colin97 commented 4 years ago

In order to train on your own dataset, you may need to:

  1. use the PCN code to generate the partial point cloud.
  2. Denoise the partial point cloud.
  3. Sample ground truth point cloud.
mactavish10 commented 4 years ago

Is it necessary to use the PCN code for generating the partial point clouds? I have written a program of my own which generates randomly shaped holes in point clouds.

Colin97 commented 4 years ago

I think it should also work if you retrain the network.

mactavish10 commented 4 years ago

My dataset isn't big enough to train the model from scratch, which is why I wanted to opt for transfer learning or fine-tuning the pre-trained model.