Learning Resources

A collection of learning resources for NeRF-or-Nothing.

Table of Contents

1. NeRF
2. Structure from motion
3. Web Servers

---

Neural Radiance Fields - NeRF

What is a NeRF?

• A Gentle Introduction
• Original NeRF project
• Great video explanation by Yannic Kilcher
• NeRF Explosion 2020
• NeRF example and code walk through in Keras
• Building a NeRF from scratch in pytorch

Current State of NeRF

Research paper Collections

• https://paperswithcode.com/method/nerf
• https://github.com/visonpon/New-View-Synthesis
• https://github.com/yenchenlin/awesome-NeRF (slightly out of date)

Individual papers

• Original paper
• NVIDIA paper with massive performance increases of 1000X over the original paper, however, it is not-fully open source software 👎
  Instant Neural Graphics Primitives with a Multiresolution Hash Encoding Explanation
• Use pretrianed Visual Transformers (Vit) to aid with estimating camera pose to create NeRFs with as little as 1-8 views vs 20-100 for the original NeRF.
  Putting NeRF on a Diet: Semantically Consistent Few-Shot View Synthesis
• Replacing the MLP (Neural Network) with a voxel representation for faster convergence.
  Direct Voxel Grid Optimization Super-fast Convergence for Radiance Fields Reconstruction

Misc NeRF projects

Deeper research paper analysis

What will be used in this project

Right now the plan is to create an API to potentially integrate multiple NeRF back ends into the web server. Research is being done into which NeRF method will be fast enough

---

Structure From Motion - SFM

---