Skip to content

Latest commit

 

History

History
213 lines (176 loc) · 8.59 KB

README.md

File metadata and controls

213 lines (176 loc) · 8.59 KB
Raw

ARCH2S: Dataset and Benchmark for Learning Exterior Architetural Strutures

ARCH2S is a semantically-enriched, photo-realistic 3D architectural models dataset and benchmark for semantic segmentation.

Our preprocess data are available and can also be downloaded by filling in the Google Form

Highlights

  • April, 2024*: ARCH2S Dataset request form is created.
  • April, 2024: ARCH2S Paper is accepted by CVPRW 2024
  • March 2024: ARCH2S repo is created; the dataset, paper, and full code will be released soon.
  • March, 2024: ARCH2S Uploads the semantic views with “Beam” and “Ceiling” labels from our ARCH2S models

Overview

TODO

  • [✔] Initial create the repo
  • [✔] dataset, benchmark and code preview for ARCH2S
  • [✔] download link for the raw data of ARCH2S dataset.
  • [✔] Paper, dataset, benchmark and full code for ARCH2S

Installation

Experiment Settings

  • Ubuntu: 22.04
  • CUDA: 11.6
  • PyTorch: 1.12.1
  • cuDNN: 7.4.1
  • GPU: Nvidia GeForce RTX 4090 x 2
  • CPU: AMD Ryzen 9 7950X 16-Core Processor @ 4.50 GHz

Conda Environment

conda create -n ARCH2S python=3.9.18 -y
conda activate ARCH2S
conda install ninja -y
# Choose version you want here: https://pytorch.org/get-started/previous-versions/
conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 cudatoolkit=11.6-c pytorch -y
conda install h5py pyyaml -c anaconda -y
conda install sharedarray tensorboard tensorboardx yapf addict einops scipy plyfile termcolor timm -c conda-forge -y
conda install pytorch-cluster pytorch-scatter pytorch-sparse -c pyg -y
pip install torch-geometric

# spconv (SparseUNet)
# refer https://github.com/traveller59/spconv
pip install spconv-cu116

# PTv1 & PTv2 or precise eval
cd libs/pointops
# usual
python setup.py install
# docker & multi GPU arch
TORCH_CUDA_ARCH_LIST="ARCH LIST" python setup.py install
# e.g. 7.5: RTX 3000; 8.0: a100 More available in: https://developer.nvidia.com/cuda-gpus
TORCH_CUDA_ARCH_LIST="7.4.1" python setup.py install
cd ../..

# Open3D
pip install open3d

Data Preparation

Our dataset and benchmark(To be released soon)

How we prepare the dataset:

  • Mining the data from the raw FBX models and convert it to point cloud data.
  • Texture mapping and colorization of the mesh with the .jpg file.
  • Sampling the point cloud (5M) data from the mesh .
  • Labeling the point cloud data with the semantic label.

Model Zoo (To be released soon)

Benchmark Results (To be released soon)

Acknowledgement

Our benchmark results implemented the following excellent works:

Model Backbone:

Model_1, Model_2, Model_3, Ours, (a), (b)

Citation

If you find this project useful in your research, please consider cite:

@misc{cheung2024arch2sdatasetbenchmarkchallenges,
      title={ARCH2S: Dataset, Benchmark and Challenges for Learning Exterior Architectural Structures from Point Clouds}, 
      author={Ka Lung Cheung and Chi Chung Lee},
      year={2024},
      eprint={2406.01337},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2406.01337}, 
}