Scholarlink

We present a crowdsourcing and mapping pipeline that enables the generation of accurate neural radiance field (NeRF) maps from camera feeds of autonomous vehicles. We generate synthetic image feeds from the CARLA driving simulator, we compare three different models (Nerfacto, TensoRF, Instant-NGP), and evaluate the accuracy of the reconstructions based on the number of acquisition vehicles and trajectory variations. We also demonstrate interactive visualization of the NeRF maps in a virtual reality headset.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Investigating Crowdsourced Neural Radiance Maps for Autonomous Vehicles

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Morui Zhu,
  • Erik Szász,
  • Mátyás Szántó,
  • Márton Vaitkus,
  • Gábor Sörös

摘要

We present a crowdsourcing and mapping pipeline that enables the generation of accurate neural radiance field (NeRF) maps from camera feeds of autonomous vehicles. We generate synthetic image feeds from the CARLA driving simulator, we compare three different models (Nerfacto, TensoRF, Instant-NGP), and evaluate the accuracy of the reconstructions based on the number of acquisition vehicles and trajectory variations. We also demonstrate interactive visualization of the NeRF maps in a virtual reality headset.