search

Yang7879 / 3D-RecGAN-extended

🔥3D-RecGAN++ in Tensorflow (TPAMI 2018)
https://arxiv.org/abs/1802.00411
MIT License
132 stars 29 forks source link

Kinnect data capture #10

Open Hans1984 opened 4 years ago

Hans1984 commented 4 years ago

Hi, your excellent work insipire me a lot. I collect some real data via the way used in the paper, however, could you please tell me the specific depth map resolution the kinnect take? Thanks a lot.

Yang7879 commented 4 years ago

hi @Hans1984 , thanks for your interest in our paper. The depth map resolution is 640x480.

Hans1984 commented 4 years ago

Thanks for your kindly reply. What's more, I find other CAD datasets, such as ModelNet, could I make use of the blender script to generate training data directly, or need some modification? Thanks a lot.

Yang7879 commented 4 years ago

hi @Hans1984 , there's no need to modify the code for rendering images and voxelization, but you need to slightly change a few lines to load the ModelNet objects which are the '.off' format (The ShapeNet objects are '.obj' format).

Hans1984 commented 4 years ago

Got it. so many thanks!