Scholarlink

Neural Radiance Fields (NeRFs) have received significant attention in reconstruction of complex 3D scenes due to their novel view synthesis capabilities. Their volumetric scene representation capabilities and flexibility in handling inherent challenges in satellite imagery distinguish NeRFs from previous approaches to satellite image analysis. Nonetheless, the evaluation of NeRF variants within the context of satellite image analysis remains limited. This study presents a comprehensive assessment of the NeRF and its variants using quantitative and qualitative metrics. We systematically evaluate and compare the performance of NeRF, Sat-NeRF, Shadow NeRF (S-NeRF), and their solar corrective variants. Our analysis explores hyperparameter tuning, rendering quality, memory utilization, and computational requirements while showing how NeRF variants tailored for satellite imagery show promise. Given the unique challenges presented by satellite imagery, this comparative study presents a thorough evaluation of various NeRF variants on an expanded dataset and offers insights into performance and efficiency trade-offs.

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

An Empirical Evaluation of the Impact of Solar Correction in NeRFs for Satellite Imagery

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Devjyoti Chakraborty,
  • Kriti Ghosh,
  • Zaki Sukma,
  • In Kee Kim,
  • Lakshmish Ramaswamy,
  • Suchendra M. Bhandarkar,
  • Deepak R. Mishra

摘要

Neural Radiance Fields (NeRFs) have received significant attention in reconstruction of complex 3D scenes due to their novel view synthesis capabilities. Their volumetric scene representation capabilities and flexibility in handling inherent challenges in satellite imagery distinguish NeRFs from previous approaches to satellite image analysis. Nonetheless, the evaluation of NeRF variants within the context of satellite image analysis remains limited. This study presents a comprehensive assessment of the NeRF and its variants using quantitative and qualitative metrics. We systematically evaluate and compare the performance of NeRF, Sat-NeRF, Shadow NeRF (S-NeRF), and their solar corrective variants. Our analysis explores hyperparameter tuning, rendering quality, memory utilization, and computational requirements while showing how NeRF variants tailored for satellite imagery show promise. Given the unique challenges presented by satellite imagery, this comparative study presents a thorough evaluation of various NeRF variants on an expanded dataset and offers insights into performance and efficiency trade-offs.