train the E-Mesh

News

• [2025.09.15] Update training E-Mesh code and release OpenLRM v1.1.1.

Setup

Installation

git clone https://github.com/Mrguanglei/Instantmesh_scriptData.git
cd OpenLRM

Environment

• Install requirements for OpenLRM first.

  conda create --name Openlrm  python=3.9 -y
  pip install -r requirements.txt
  

• Please then follow the xFormers installation guide to enable memory efficient attention inside DINOv2 encoder.

Quick Start

Dataset format

|--rendering_random_32views
|--- 1 # object id
| |--- 000.png
| |--- 000_normal.png
| |--- 000_depth.png
| |--- ...
| |--- 031.png
| |--- 031_normal.png
| |--- 031_depth.png
| |--- camera.npz # 
|--- 2 # object id
| |--- 000.png
| |--- 000_normal.png
| |--- 000_depth.png
| |--- ...
| |--- 031.png
| |--- 031_normal.png
| |--- 031_depth.png
| |--- camera.npz #
|--- 3 # object id
| |--- 000.png
| |--- 000_normal.png
| |--- 000_depth.png
| |--- ...
| |--- 031.png
| |--- 031_normal.png
| |--- 031_depth.png
| |--- camera.npz # 
|-- valid_paths.json
|-- ...

valid_paths.json fomerat:

{
  "good_objs":[
      "1",
      "2",
      ...
  ]

}

Downloading the dataset

I took over 200 glb files from the Objaverse dataset and used the glb to render the dataset we needed

1.Downloading the dataset:Dataset address.

2.Place the glb file in the data folder

Modifying the script

• Find the script that we need to modify scripts/data/objaverse/blender.sh,

  DIRECTORY="/your/path/OpenLRM/data"    #Put the path to the dataset file you downloaded here

Blender:

Run blender.sh .

This will automatically render the dataset we need above.

Tips

• The recommended PyTorch version is >=2.1. Code is developed and tested under PyTorch 2.1.2.
• If you encounter CUDA OOM issues, please try to reduce the frame_size in the inference configs.
• You should be able to see UserWarning: xFormers is available if xFormers is actually working.
• If there is no module in bpy and mathutils, please look up the information yourself.

Citation

If you find this work useful for your research, please consider citing:

@article{hong2023lrm,
  title={Lrm: Large reconstruction model for single image to 3d},
  author={Hong, Yicong and Zhang, Kai and Gu, Jiuxiang and Bi, Sai and Zhou, Yang and Liu, Difan and Liu, Feng and Sunkavalli, Kalyan and Bui, Trung and Tan, Hao},
  journal={arXiv preprint arXiv:2311.04400},
  year={2023}
}

License

• OpenLRM as a whole is licensed under the Apache License, Version 2.0, while certain components are covered by NVIDIA's proprietary license. Users are responsible for complying with the respective licensing terms of each component.

• Model weights are licensed under the Creative Commons Attribution-NonCommercial 4.0 International License. They are provided for research purposes only, and CANNOT be used commercially.

  Star History