Neural-Point-Cloud-Rendering-via-Multi-Plane-Projection
Neural-Point-Cloud-Rendering-via-Multi-Plane-Projection copied to clipboard
Is RGB pixel feature necessary?
For some colorless point cloud files, if there is no RGB pixel feature as the model input, but it is used to calculate the perceived loss, can the model work properly?
I haven't tried training this network without RGB initialization yet. But I think it's possible as shown in this project https://github.com/alievk/npbg.
@daipengwa Thank you for your reply. Could you share me the pre-trained model weight of neural render net?
This is one pre-trained model weight on ScanNet 'scene_0000_00', https://drive.google.com/drive/folders/1rP7xDhOtx-L24ymJGY3A0EgxSDVEAAAA?usp=sharing.