NVlabs / DeepIM-PyTorch

PyTorch implementation of the DeepIM framework
Other
91 stars 19 forks source link

How to Render YCB Video dataset based custom data, if its textureless? #12

Open sag-Nik opened 1 year ago

sag-Nik commented 1 year ago

FILE: ycb_video.py (in lib>datasets) LINE ~ 77,78

3D model paths

self.model_mesh_paths = ['{}/{}/textured_simple.obj'.format(self._model_path, cls) for cls in self._classes_all] self.model_texture_paths = ['{}/{}/texture_map.png'.format(self._model_path, cls) for cls in self._classes_all]

This is asking me to give texture files for the rendering purpose, but I am working on textureless data of YCB Video format generated by Isaac Sim. Is there a work around?

chengm0-0 commented 12 months ago

Hi, have you solved this question?

sag-Nik commented 12 months ago

Nope. It was a bit difficult for me to work with my custom object bcz the algorithm requires PoseCNN results to start with. So I altogether quit trying this framework.

You could try using a fake texture file, or generate a texture file for your object using Blender maybe. I was going to try those.