Hello kwea, sorry for disturbing you! I would like to ask a question.
I noticed that the fine stage merged the sample points from the coarse stage and sent them to the network for inference, here I have a doubt, because the sample points from the coarse stage have already been inferred before, why do they have to be sent to the network for inference again here, is it to facilitate the final volume rendering? But it feels like this will add a lot of computation
https://github.com/kwea123/nerf_pl/blob/52aeb387da64a9ad9a0f914ea9b049ffc598b20c/models/rendering.py#L229
Hello kwea, sorry for disturbing you! I would like to ask a question.
I noticed that the fine stage merged the sample points from the coarse stage and sent them to the network for inference, here I have a doubt, because the sample points from the coarse stage have already been inferred before, why do they have to be sent to the network for inference again here, is it to facilitate the final volume rendering? But it feels like this will add a lot of computation https://github.com/kwea123/nerf_pl/blob/52aeb387da64a9ad9a0f914ea9b049ffc598b20c/models/rendering.py#L229