Closed pansanity666 closed 7 months ago
Hi, the kernel size controls the the size of our 2D mip filter and it is choose to approximate a single pixel (box filter). The following figure shows Gaussians with different kernel sizes (0.1, 0.3, and 1.0) and a box filter that we want to approximate. Using kernel size = 0.1 approximates the box filter well.
Thanks for your clear explanation!
Hi,
Thanks for your great work! A stupid question is that what exactly does kernel_size in the code mean, and why it is set as 0.1 for all the exps?
Best,