OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Nerf Mae

Nerf Mae is PyTorch research code for masked autoencoders that learn self-supervised 3D representations from neural radiance fields. The repository includes model checkpoints, radiance-and-density-grid datasets, and code for downstream 3D object-detection experiments.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Nerf Mae — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Nerf Mae when a research workflow needs learned features from neural radiance field grids, or when exploring masked-autoencoder pretraining and 3D detection with the released code and data.

Who it is for

  • Researchers studying self-supervised 3D representation learning or neural radiance fields.
  • Developers building downstream 3D perception experiments that can use NeRF-derived features.

Documented uses

  • Extract learned feature maps from NeRF scene grids for use in a downstream task.
  • Use a pretrained or finetuned checkpoint as a starting point without conducting large-scale pretraining.
  • Explore pretraining, fine-tuning or evaluation workflows for 3D object detection on NeRF data.
Before you choose

Requirements & limits

The code repository is public and licensed under MIT. The README links to pretrained and finetuned checkpoints and downloadable datasets.

  • The documented environment uses Python and PyTorch 1.12.1 with CUDA 11.3; the authors say the code was built and tested on CUDA 11.3.
  • Downstream fine-tuning requires compiling a CUDA extension.
  • The released data is organized around NeRF radiance and density grids rather than ordinary image inputs.
  • The documented downstream detection example uses NeRF-RPN as a 3D object-detection head.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up the NeRF-MAE Python environment

Prepare the documented Python environment and install the project requirements and CUDA 11.3 PyTorch packages.

Before you start

  • A local NeRF-MAE repository checkout.
  • Conda and pip.
  • The README specifies a Python 3.9 Conda environment and says the code was built and tested on CUDA 11.3.

Steps

  1. From the repository, create and activate the Conda environment with conda create -n nerf-mae python=3.9 and conda activate nerf-mae.

  2. Upgrade pip with the command below.

    Documented command
    pip install --upgrade pip

  3. Install project dependencies with the command below.

    Documented command
    pip install -r requirements.txt

  4. Install the specified PyTorch packages with the command below.

    Documented command
    pip install torch==1.12.1+cu113 torchvision==0.13.1+cu113 -f https://download.pytorch.org/whl/torch_stable.html

Expected outcome

The README's Python environment and specified PyTorch packages are installed.

Guide limitations

  • The README says the code was built and tested on CUDA 11.3; it does not document support for other CUDA versions or operating systems.
  • The README's opening sentence says Python 3.7, while its environment commands create Python 3.9; the commands specify Python 3.9.
  • This setup does not cover running pretraining, finetuning, or inference.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
zubair-irshad/NeRF-MAE · Official homepage title · retained in its source language
How we review sourcesSuggest a correction