FacadeNet

FacadeNet's code is built based on the swapping-autoencoder repo code structure

Conda Environment

Create new conda environment

conda env create -f environment.yml

Datasets

Structure

Root

• Facades(Rectified facade images downloaded from lsaa
• horizontal maps
• vertical maps
• DINO features(Extract features using DINOv2)
• Depth

Click on the link to download the data: Link

Training

Change data file paths in config file (experiments/facades_vec_pc_lr_launcher.py)

Run

python -m experiments facades_vec_pc_lr train facadenet

Testing

Change data file paths in config file (experiments/facades_vec_pc_lr_launcher.py)

Create folder named checkpoints/facadenet/

Download facadenet pre-trained weights Link in checkpoints/facadenet/ folder

Run

python -m experiments facades_vec_pc_lr test facadenet