search

xiumingzhang / GenRe-ShapeHD

Code and Data Release for GenRe (NeurIPS 2018) and ShapeHD (ECCV 2018)
http://genre.csail.mit.edu/
294 stars 62 forks source link

Training GenRe with my own dataset - how to generate it #1

Closed vicentepedro closed 5 years ago

vicentepedro commented 5 years ago

Hi Xiuming!

Thanks for sharing the code. I would like to train GenRe on my own dataset.

I already have:

  1. the RGB image
  2. the silhouette image
  3. the normals
  4. the aligned 3D model (voxel representation)

I need, at least, the spherical maps as well, right?

How can I create my own dataset to train GenRe? Do you have any example on how you created yours on chairs, aeroplanes and cars based on shapenet?

Thanks in advance, Best, Pedro

xiumingzhang commented 5 years ago

Hi Pedro,

Yes, you'd need to generate spherical maps from depth maps. For this purpose, you should use this class: https://github.com/xiumingzhang/GenRe-ShapeHD/blob/master/toolbox/spherical_proj.py#L31.

For creating your own dataset (and getting a sense of what data you need to train GenRe), we recommend taking a look at the released data that we used to train GenRe (please see https://github.com/xiumingzhang/GenRe-ShapeHD#data).