Closed Ailuridae closed 5 years ago
This repo contains a PyTorch reimplementation of the original repo. It should be more appropriate to ask this question in the original repo. I understand what you are saying, and feel free to implement a different sampling method and submit a PR if you want.
The original paper states:
This is not correct, as meshes with large faces and detailed textures will result in a blurry images. The problem, which was already described in #6, could probably be solved by rewriting the current texture sampling method. What's the reason behind the current implementation?