bmild
why near: 2 and far: 6
In the article it is said "For experiments with synthetic images, we scale the scene so that it lies within a cube of side length 2 centered at the origin, and only query the representation within this bounding volume", and in the code the near is set to 2 and the far is set to 6. Feel confused about that. Thank you :-)
It depends on which dataset you are using. 4 and 6 are for datasets from blender (load-blender.py file) while the load_llff.py file uses another method to calculate near and far amounts. On the other hand, [-1,1] cube is for normalizing datasets along with each axis (x,y, z). Indeed, they bring the dataset into limited intervals (cubes) to have the simplest process. In summary, [-1.1] normalized cube method is not the same near and far method. You can eliminate only near and far concepts in particular datasets.
