octree-nn / ocnn-pytorch

Octree-based 3D Convolutional Neural Networks
MIT License
156 stars 17 forks source link

key advantages of key advantages #2

Closed samanemami closed 2 years ago

samanemami commented 2 years ago

What are the key advantages of your library in comparison to the microsoft/O-CNN? I could not find any related description in the documentation that tells the difference between these two libraries.

Thank you

wang-ps commented 2 years ago

octree-nn/ocnn-pytorch is based on pure PyTorch, while the core of microsoft/O-CNN is based on C++/CUDA. Therefore, octree-nn/ocnn-pytorch is easier to build and more compatible.

samanemami commented 2 years ago

Thank you.

Yes, indeed is easier to build and compatible with different OS. So, the main performance (In terms of building sparse convolutional layers) of both is the same?

wang-ps commented 2 years ago

The functionality is the same. The memory consumption is similar, and as for speed, octree-nn/ocnn-pytorch runs even faster.

samanemami commented 2 years ago

Perfect, thanks