Multiscale Representation for Real-Time Anti-Aliasing Neural Rendering

This repository contains the implementation of Mip-VoG described in D. Hu, Z. Zhang, T. Hou, T. Liu, H. Fu, M. Gong: Multiscale Representation for Real-Time Anti-Aliasing Neural Rendering. International Conference on Computer Vision (ICCV) 2023.

Training and evaluation with different configuration

To train a single scene with mip-vog, run python run.py --config configs/nerf_ms/"$scene".py --resolution 512 --mip_train

Test the model with mip-vog, run python run.py --config configs/nerf_ms/"$scene".py --resolution 512 --mip_train --render_only --render_test --mip_test

Test the model with voxel grids but without mipmapping, run python run.py --config configs/nerf_ms/"$scene".py --resolution 512 --mip_train --render_only --render_test

Citation

If you find it useful, please consider citing:

@InProceedings{Hu_2023_ICCV,
    author    = {Hu, Dongting and Zhang, Zhenkai and Hou, Tingbo and Liu, Tongliang and Fu, Huan and Gong, Mingming},
    title     = {Multiscale Representation for Real-Time Anti-Aliasing Neural Rendering},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2023},
    pages     = {17772-17783}
}

Acknowledgement

The training code base is origined from DVGO, and web viewer base is origined from SNeRG.