BuildingBlock

[SIGGRAPH 2025]BuildingBlock: A Hybrid Approach for Structured Building Generation

Installation

1. Download Docker

Ensure Docker is properly installed on your system. Visit Docker’s official website for installation instructions specific to your operating system.

docker pull dreaminghuang/building_block:0.1

2. Run Docker Container

Mount your local directory to the container and start an interactive session:

docker run -it --gpus all -v your_path_of_building_block:/building_block -w /building_block building_block:0.1 /bin/bash

Replace your_path_of_building_block with the absolute path to your local directory where you want to store the project files.

3. Initialize Environment

Run the initialization script to set up the required dependencies:

bash initialization.sh

Dataset Preparation

1. Dataset Download and Extraction

Download the dataset from
👉 https://huggingface.co/datasets/dreaming-huang/buildingblock/blob/main/building_block_data_opensource_only_layout_and_cond.zip
Extract the BoxCenterSizeLabel_all directory from the building_block_data_opensource_only_layout_and_cond.zip:

unzip building_block_data_opensource_only_layout_and_cond.zip

2. Process Dataset

Run the following Python scripts in sequence to preprocess the data:

python 1-json_rotate_augment.py

This script performs data augmentation through rotation of the original JSON files.

python 2-normUeJson.py

This script normalizes the Unreal Engine JSON format data.

python 3-json2boxnp.py

This script converts the JSON data to NumPy arrays in box representation format.

cp dataset_stats.txt ./BoxCenterSizeLabelNp

Running the Model

Navigate to the scripts directory:

cd scripts

Execute the commands provided in command.sh to train and/or evaluate the model:

# View the available commands
cat command.sh

# Execute specific commands as needed
# For example:
python train_diffusion_building_DDP.py ../config/text/diffusion_building_DIT.yaml uncond  --experiment_tag uncond --n_processes 0 --with_swanlab_logger

Project Structure

• 1-json_rotate_augment.py: Data augmentation script
• 2-normUeJson.py: JSON normalization script
• 3-json2boxnp.py: JSON to NumPy conversion script
• scripts/: Contains model training and evaluation scripts
• configs/: Configuration files for different model settings
• BoxCenterSizeLabel_all/: Directory containing the dataset

Notes

• Make sure your GPU drivers are properly configured for Docker GPU passthrough
• The dataset processing may take significant time depending on the size of the dataset
• Check the log files for any errors during processing

For more detailed information about the model architecture and training parameters, please refer to the documentation in the respective script files.