ngp_lidar.md

[neuralsim] InstantNGP + UrbanNeRF

Instant-NGP | UrbanNeRF

⚠️ Unofficial implementation ⚠️

An unofficial implementation (combination) of "Instant Neural Graphics Primitives with a Multiresolution Hash Encoding" and "Urban Radiance Fields"

@article{muller2022instantngp,
	title={Instant neural graphics primitives with a multiresolution hash encoding},
	author={M{\"u}ller, Thomas and Evans, Alex and Schied, Christoph and Keller, Alexander},
	journal={ACM Transactions on Graphics (ToG)},
	volume={41},
	number={4},
	pages={1--15},
	year={2022}
}

@inproceedings{rematas2022urban,
	title={Urban radiance fields},
	author={Rematas, Konstantinos and Liu, Andrew and Srinivasan, Pratul P and Barron, Jonathan T and Tagliasacchi, Andrea and Funkhouser, Thomas and Ferrari, Vittorio},
	booktitle={Proc. {IEEE/CVF} Conference on Computer Vision and Pattern Recognition (CVPR)},
	pages={12932--12942},
	year={2022}
}

Usage

Requirements

• ~45 mins training time on single RTX3090

• ~11 GiB GPU Mem

• >20 GiB CPU Mem (Caching data to speed up)

Dataset preparation

• Waymo Open Dataset - Perception

  • README

  • split file: waymo_static_32.lst

Major settings

Settings	Config file
Multi-view reconstruction with LiDAR	ngp_withlidar.230814.yaml

Instructions

For detailed instructions, please refer to the general guide section in code_single.