search

NVlabs / instant-ngp

Instant neural graphics primitives: lightning fast NeRF and more
https://nvlabs.github.io/instant-ngp
Other
15.82k stars 1.9k forks source link

confused on how this works #1138

Open mihailpettas opened 1 year ago

mihailpettas commented 1 year ago

is this a 2d to 3d converter model? or do i need to train it for every object?

jc211 commented 1 year ago

This converts posed images (images where the position of the camera that took them is also known) into a 3D "model". What you're seeing is a volumetric field being rendered. That field can be converted to a semi usable 3D mesh if the training goes OK. What is being constructed is whatever is in the images. If your images only have one object in them, then the model is only of that object.