This is the officially implemented of paper "SG-NeRF: Semantic-guided Point-based Neural Radiance Fields"
Attention!!! 相机插值功能基于 mitsuba2 开发,只能是mitsuba2,不能是mitsuba3
要求 gcc>=8.4 步骤:
https://mitsuba2.readthedocs.io/en/latest/src/getting_started/compiling.html
`bash export CC=clang-9 export CXX=clang++-9
mkdir build cd build cmake -GNinja .. ninja
#然后将接下来的代码加入~/.bashrc MITSUBA_DIR=/home/vr717/Documents/qys/code/mitsuba2 MITSUBA_BUILD_DIR="build" export PYTHONPATH="$MITSUBA_DIR/dist/python:$MITSUBA_DIR/$MITSUBA_BUILD_DIR/dist/python:$PYTHONPATH" export PATH="$MITSUBA_DIR/dist:$MITSUBA_DIR/$MITSUBA_BUILD_DIR/dist:$PATH" `
如果不需要相机插值,只需要注释掉前面的,并注释掉报错的方法
python import mitsuba mitsuba.set_variant('scalar_rgb') from mitsuba.core import ScalarTransform4f, AnimatedTransform
Code is based on awesome Point-NeRF.
@article{xu2022point,
title={Point-NeRF: Point-based Neural Radiance Fields},
author={Xu, Qiangeng and Xu, Zexiang and Philip, Julien and Bi, Sai and Shu, Zhixin and Sunkavalli, Kalyan and Neumann, Ulrich},
journal={arXiv preprint arXiv:2201.08845},
year={2022}
}