how to set kernel_size

[liwenssss]

Hi, thanks for your excellent work! I noticed that in your rendering process, you add a kernel_size parameter. However, I couldn't find any issue regarding its settings. It seems that the kernel_size is set to 0. Could you please explain the purpose of this parameter and how I should adjust it based on experimental results?

[BaowenZ]

Hi, it is 2D filter size. Since we did not use it in our experiments, we have set it to 0. You could use it to address aliasing caused by low resolution or small 2D kernel size.

[liwenssss]

Thanks for your reply. Another question, I want to save the rendered normal map. I first normalize the rendered_normal to the [0, 1] range and then use torchvision.utils.save_image to save it. However, the saved result appears yellow or even red. Is this normal?

[BaowenZ]

Hi, we show the normal by:

plt.imshow((-render_pkg["normal"].detach().cpu().numpy().transpose(1,2,0)+1)/2)

[liwenssss]

Hi, I have another question. I want to use your module to get normal vector, But I dont know how to do. Is it any possible to achieve it by your module?